Task Planning Research Articles

BiCap: A novel bi-modal dataset of daily living dual-arm manipulation actions

Dual-arm manipulation of daily living objects is essential for robots operating in household environments. Learning from demonstration is a promising approach for teaching robots dual-arm manipulation skills. It usually requires a dataset of participants demonstrating the sequence of manipulation actions to achieve a goal. That sequence can be represented using symbolical or trajectory encoding. The symbolic-encoded sequence is known as the task plan. The chief limitations of current datasets are that most tend to disregard dual-arm manipulation skills and omit the formal grammar used to annotate the task plans. This paper introduces BiCap, a novel bi-modal dataset of dual-arm manipulation actions on daily living objects coupled with a bio-inspired action context-free grammar for fine-grained task plan annotation to train, test, and validate learning from demonstration-based algorithms for robotic dual-arm manipulation. 15 participants were recruited. The experimenter placed reflective markers on their upper limbs and pelvis. Then, the participants sat at a table where one or two objects were placed. They performed one of the following tasks: pouring, opening, and passing, using both hands. An RGB camera pointing towards the table recorded the participants’ hand movements. Subsequently, an annotator reviewed the RGB videos and wrote the participants’ task plans using the bio-inspired action context-free grammar. The participants’ upper-limb kinematics were computed, too, which provides the trajectory-encoded action sequences. The resulting dataset, BiCap, contains 4,026 task plans, videos, and motion data of the 15 participants.

Responding to AI

The influence of artificial intelligence “has enormous implications for higher education” (Bearman, Ryan & Ajjawi, 2023, p. 369). Those of us teaching in the higher education sector have felt this profoundly: after the rapid uplift in digital teaching and learning practices in response to COVID 19 we are required, yet again, to respond to a new set of challenges brought about by the impact of AI. As an example, chatbot technology is calling into question the reliability of the traditional essay as a valid assessment tool (Stokel-Walker, 2022), requiring significant critical reflection and reevaluation of assessments from teaching academics to ensure course learning outcomes are met. This presentation outlines how a teaching specialist academic working in an initial teacher education (ITE) programme at a South Australian university carried out one such redesign. The focus of the re-design was a ‘unit/year plan’ assessment task for a final year curriculum specialization topic. The redesign aligned with the two guiding principles from TEQSA’s Assessment reform for the age of artificial intelligence (Lodge et al., 2023). The first of these guiding principles: ‘Assessment and learning experiences equip students to participate in a world where AI is ubiquitous’ (p. 2) points to the importance of educating students around AI use relevant to their learning. The second guiding principle: ‘Forming trustworthy judgements about student learning in a time of AI requires multiple, inclusive and contextualised approaches to assessment’ (p. 3) underscores the ongoing necessity for universities to have effective means by which to measure student learning outcomes, especially in accredited courses. The redesign of the assessment task both acknowledged the potential of AI for use in the professional space, and simultaneously sought to ensure assessment validity through the inclusion of an interview with the pre-service teachers. Overall, the redesign of the assessment task incorporated three components. Firstly, a brief introductory module about discipline-specific uses of AI was incorporated into the online course content. Secondly, pre-service teachers were ‘permitted’ to utilise available AI platforms in the creation of their planning task. Thirdly, the assessment process was adapted to include an oral presentation (Gulikers, Bastiaens & Kirschner, 2004) conducted as a panel interview in which each pre-service teacher could provide evidence of the pedagogical perspectives underpinning their plan and subsequently demonstrate the successful achievement of topic learning outcomes. The purpose of the panel interview was to assist assessors in forming a judgement about pre-service teachers’ pedagogical understanding and application, in line with relevant criteria from Matheis and John’s Comprehensive Authentic Assessment Model (CAAM) (Matheis & John, 2024). Autoethnographic reflections on the first delivery of the redesigned assessment are provided, including alternative approaches for future iterations. The learnings from the teaching initiative outlined in this presentation contribute to the growing discussion about how we can develop the capacity of our pre-service teachers as skilled and critical practitioners who can utilize AI as a valuable tool for their ongoing professional practice.

A Hierarchical Planning Method for AUV Search Tasks Based on the Snake Optimization Algorithm

In a complex and dynamic battlefield environment, enabling autonomous underwater vehicles (AUVs) to reach dynamic targets in the shortest possible time using global autonomous planning is a key issue affecting the completion of search tasks. In this study, ahierarchicalAUV task planning method that uses a combination of hierarchical programming and a snake optimization algorithm is proposed for two typical cases where the platform can provide initial target information. This method decomposes the search task problem into a three-level programming problem, with the outer task planning goal of achieving the shortest encounter time between AUV and dynamic targets; the goal of task planning in the middle layer is to achieve the shortest actual navigation time for AUVs under different operating conditions; and the internal task planning is responsible for considering the comprehensive trajectory optimization under navigation constraints such as threat zone, path length, and path smoothness. The snake optimization algorithm was used for solving each layer of task planning. The feasibility of the proposed method was verified through simulation experiments of AUV search tasks under two types of initial target information conditions. The simulation results show that this method can achieve task planning for AUV searching for dynamic targets under various constraint conditions, optimize the encounter time between AUV and dynamic targets, and have strong engineering practical value. It has certain reference significance for task planning problems similar to underwater unmanned equipment.

Classifying Technologies during the Assessment, Treatment Planning, Documentation and Evaluation Phases of Music Therapy: A Survey of Board-Certified Practitioners

Music therapists (MT-BCs) use diverse technologies to provide evidence-based personalized interventions to a wide variety of people. Most studies on the technological practices of MT-BCs report a general overview of the tools they use in their daily work. This study offers an new way of understanding technologies used by MT-BCs, classified and situated in the phases of music therapy (MT) practice: Assessment, Treatment Planning, In-Session, Documentation and Evaluation. An online survey was sent to a mailing list of 1,951 board-certified music therapists (MT-BC), and we received 104 responses. Results support distinct functions in technological practices between each of the phases of MT, revealing categories of notetaking and data entry that characterize planning, documentation and evaluation tasks, and a wider diversity of technology configurations in assessment and in-session work. We end by discussing design implications for HCI researchers and designers of music technologies for health, as well as HCI-MT design collaboration to better support the work of the MT-BC community.

Multirobot Assembly Sequence Planning of a Large Space Telescope

As space structures trend toward larger scales and modularization, their on-orbit assembly presents a significant challenge. Assembly sequence planning is crucial in minimizing vibration and enhancing the efficiency of on-orbit construction. This paper explores the multirobot assembly sequence planning algorithm for a large space telescope. A dynamic model is firstly developed to analyze the vibration of segmented mirror pieces and the main satellite during assembly, taking into account the flexibility of connections at interfaces. The dynamic model is able to provide criteria for time efficiency and satellite attitude disturbances. Considering constraints during the assembly process, a modified multiobjective ant colony optimization algorithm capable of planning multirobot assembly sequences is implemented. The simulation results demonstrate that the proposed method not only facilitates parallel planning of multirobot assembly sequences but also leads to a reduction in assembly time and attitude disturbances of the main satellite. The results of assembly sequence planning provide dependable guidance for the subsequent planning levels, such as task planning and trajectory planning, of the on-orbit assembly task for large space telescopes.

Characterizing the Impact of Emergent Technologies on Earth Communications Reliance for Crewed Deep Space Missions

Future human expeditions into deep space will encounter unique design challenges posed by the increasing distance from Earth, one of which is the inability to maintain near-continuous communication with ground support teams. As a result, the habitat and crew will require a higher level of operational self-sufficiency informed by onboard self-awareness and decision-making capabilities to accomplish functions that have historically involved extensive support from mission control personnel. These include, for example, task planning and anomaly detection, diagnosis and correction, as well as monitoring of consumable usage rates and system performance. Novel emergent technologies in the domain of ‘smart’ systems and digital twins can be employed to enable the onboard capabilities needed to function with minimal Earth communication. However, these technologies are generally low-TRL and difficult to incorporate through traditional systems engineering methods.This work proposes an extension of utility theory to assess the potential reduction in Earth communication reliance between baseline and emergent technologies, scored through a summation of weighted attributes and evaluated against future deep space mission needs and constraints. Case studies are defined from technology under development in the NASA Habitats Optimized for Missions of Exploration (HOME) Space Technology Research Institute (STRI), a project charged with developing and evaluating technologies to enable highly autonomous deep space habitats, and compared to the current baseline technologies used onboard the International Space Station. Notional deep space mission needs and constraints are informed from current NASA plans for the Moon and Mars. These in turn are broken out into higher resolution attributes that are used to assess the operational capabilities of each emergent technology and whether it offers a potentially viable solution for deep space. These technology attributes are scored through a series of semi-structured subject-matter interviews and compared to notional deep space mission constraints. Additional considerations for technology maturation and integration are also discussed. The design analysis paradigm described and demonstrated here can be adapted to assess the degree to which other emergent technologies rely on Earth-based communication, and is intended to provide trade space considerations for future research and development towards deep space habitat self-sufficiency.

THE DETECTION OF DISTOMOLAR TEETH ON PANORAMIC RADIOGRAPHS USING DIFFERENT ARTIFICIAL INTELLIGENCE MODELS.

One notable anomaly, presence of distomolars, arises beyond the typical sequence of the human dental system. In this study, convolutional neural networks (CNNs) based machine learning methods were employed to classify distomolar tooth existence using panoramic radiography (PR). PRs dataset, composed of 117 subjects with distomolar teeth and 146 subjects without distomolar teeth, was constructed. These images were assessed using AlexNet, DarkNet, DenseNet, EfficientNet, GoogLeNet, ResNet, MobileNet, NasNet-Mobile, VGG, and XceptionNet frameworks for distomolar teeth existence. Considering the moderate number dataset samples, transfer learning was also utilized to improve the performance of these CNN based networks along with 5-fold cross-validation. The final classification was obtained through the fusion of the classifiers results. Performance of the experimental studies was assessed utilizing accuracy (Acc), sensitivity (sen), specificity (spe) and precision (pre) metrics. Best accuracy value of 96.2% was obtained for the fusion of DarkNet, DenseNet, and ResNet, three best individual performing architectures, in distomolar teeth classification problem. In summary, this study has demonstrated the significant potential of CNNs in accurately detecting distomolar teeth in dental radiographs, a critical task for clinical diagnosis and treatment planning. The fusion of CNN architectures, particularly ResNet, Darknet, and DenseNet, has shown exceptional performance, pointing towards the future of artificial intelligence (AI) driven dental diagnostics. Our findings showed that these systems can help clinicians during radiologic examinations.

Making the Invisible, Visible: An Exploration of Track-and-Field Coaches' Perspectives of Their Planning Processes.

The traditional understanding of how coaches plan for sporting performance is rooted in the assumption that coaches create periodized plans underpinned by physiological principles, thereby providing scientific credibility to their work. In contrast, there remains a paucity of literature exploring how coaches understand or think about their planning practices. The purpose of this study was to generate new knowledge regarding what information coaches actually consider within their planning processes and how they actually approach the task of planning. Using rigorous, in-depth interviews, this study examined athletics coaches' understandings of their everyday planning practices, in an attempt to contribute to narrowing the gap between academic research and real-world application. Twenty-eight highly experienced, high-performance track-and-field coaches based in England (female, n = 1; male, n = 27) were recruited. The coaches were from the throwing disciplines (n = 10) and endurance running (n = 18). Coaches were interviewed about their planning process, using a maximum of 3 semistructured interviews per coach, spaced across a full athletics season. In total, this generated 68hours of data. The analysis demonstrated that, while the participant coaches utilized the principles of periodization, their planning activities were not limited to this issue. The findings highlighted how the coaches conceptualized successful athletic performance in a holistic way; that is, planning is multifaceted in nature. This study presents a holistic picture of the complexity of coaches' planning, detailing the considerable time and attention given to planning for athletes' psychological, personal, and social development, to enhance athlete performance and development.

Efficient Reinforcement Learning of Task Planners for Robotic Palletization Through Iterative Action Masking Learning

Efficient Reinforcement Learning of Task Planners for Robotic Palletization Through Iterative Action Masking Learning

Research on trajectory planning based on deep reinforcement learning

Abstract For trajectory planning in a three-dimensional continuous space with multiple complex constraints, traditional optimal control methods face significant challenges, such as long computation times and poor dynamic performance. Using deep reinforcement learning algorithms for training agents in flight trajectory planning can effectively improve the dynamic performance of the planning algorithms. This paper focuses on training agents for UAV emergency landing and obstacle avoidance trajectory planning based on deep reinforcement learning algorithms. Within the reinforcement learning framework, we design the training environment, state space, action space, reward function, and neural network structure. To address common issues in reinforcement learning, such as poor generalization and slow training speed, we improve the training process using the concept of curriculum learning. Comparisons with pseudospectral methods demonstrate that deep reinforcement learning can effectively handle trajectory planning tasks in real flight environments, achieving trajectory quality close to that of pseudospectral methods while offering superior real-time performance.

Path Planning in Payment Channel Networks with Multi-Party Channels

Payment Channel Networks (PCNs) provide a means to improve the scaling of cryptocurrency payments by allowing peers to make payments between themselves in an efficient manner. To make a payment between two peers, the task of path planning must first be performed to determine a path in the PCN connecting the peers in question before the payment in question is performed using this path. To date, existing research has focused on the problem of performing path planning in PCNs that contain two-party channels. It has been hypothesised that the scaling of PCNs could be further improved by considering the inclusion of multi-party channels that contain more than two peers. However, the problem of performing path planning in PCNs that contain multi-party channels has not yet been considered. In this article, we address this gap in the research literature and propose a novel path planning method for PCNs containing multi-party channels. This method involves modelling the PCN with multi-party channels as a hypergraph, a type of graph where edges can contain two or more vertices, and using this model to solve the path planning problem in question. We prove that the proposed method is correct and computationally efficient. Furthermore, assuming path planning is performed using this method, we also present theoretical and experimental analyses that demonstrate the scaling benefits of using multi-party channels.

The meaning of organizational culture and structure for interprofessional collaboration

Abstract In Germany, early neurological rehabilitation (ENR) acts as a link between acute inpatient care and rehabilitation, aiming to restore patients’ ability to rehabilitate amid their complex clinical conditions. Interprofessional cooperation (IPC) is crucial for harmonized treatment due to patients’ vulnerability. An interprofessional team comprising medical, nursing, and various therapy professionals is essential for this purpose. To explore the meaning of organizational culture and structure to IPC, 15 group discussions in 5 clinics were conducted with n = 76 participants. 1. Weekly team meetings are used to coordinate treatment goals and outcomes, but their implementation varies from clinic to clinic. Two types of meetings were identified: those that focus primarily on billing compliance and others that address patients’ long-term perspectives. However, both require additional social service involvement and are not sufficient to adequately address daily tasks, requiring additional IPC coordination. Some clinics have therefore introduced daily meetings to improve the coordination of IPC tasks. 2. Therapy planning and therapist assignment shape IPC and task fulfillment. Three therapy planning methods were observed: centralized, block-based, and flexible, each with different therapist allocation strategies. Centralized planning with flexible allocation leads to appointment pressure and interprofessional communication challenges. Block therapy reduces therapist turnover and promotes better interprofessional coordination in patient care. Flexible therapy planning is conducted through daily interprofessional meetings with shared task allocation and planning leads to a shared understanding. Initial analysis underscores the meaning of organizational structure on interprofessional task fulfillment in ENR, emphasizing the importance of enhancing different cooperation formats which can balance stability and flexibility and integrate social services effectively in patientcare. Key messages • Morning interprofessional meetings on the ward facilitate flexible responses to daily changes in patients’ health conditions and promote interprofessional collaboration. • The consideration of occupational group-specific processes in ward organization is a necessary structural prerequisite for successful interprofessional collaboration.

Examinations of search interactions, experiences and communication in collaborative search

PurposeThis study aims to explore the characteristics of collaborative information searching by comparing the differences in search experience and interaction between collaborative search mode and individual search mode and by analyzing the communication content and patterns during collaborative searches.Design/methodology/approachA user experiment was conducted using the Conference and Labs of the Evaluation Forum (CLEF)-Social Book Search platform, recruiting 16 individual and 18 collaborative participants. Each of the participants was required to complete two types of book search tasks in a lab setting. Interactions with the system were logged, and participants' experiences were captured through pre/post-search questionnaires. Additionally, the communication among collaborative participants was recorded and coded.FindingsThis study identified characteristics of collaborative information searching in three aspects. (1) Search experience: collaborative searchers showed a higher appreciation for the system’s functionality and aesthetics and were more engaged than individual searchers, despite experiencing slightly more mental strain. (2) Search strategies: in focus tasks, collaborative searchers made more decisions than individual searchers, indicating a need for book sorting. In open tasks, collaborative searchers take longer before making decisions, likely due to a higher level of interest and participation. (3) Communication content and pattern: collaborative searchers mainly discussed understanding and judgment of books and task plan. Searchers with same cognitive style talked more about search operations than those with different cognitive styles, while less about book judgment. Three communication patterns were identified: understanding-judgment, judgment-dominant and balanced. Judgment-dominant participants generally reported higher satisfaction with their search results than those in the other two patterns.Originality/valueThis study enhances understanding of collaborative search by comparing it with individual search and analyzing communication content, potentially inspiring collaborative search system and experiment design. Additionally, the coding table for communication between collaborative searchers offers valuable insights for future research.

MANAGING PERSONAL SELLING ACTIVITIES IN A MODERN ENTERPRISE

Proper management is necessary for the selling function to survive effectively and efficiently in the organizational structure of an enterprise. By managing sales, i.e. sales force, the enterprise strives to accomplish the ‘-proclaimed’ goals and strategy. Within organizational sales, personal selling activities are also subordinated to managerial competencies (managers, directors, etc.), which are hierarchically positioned and have formal responsibilities. This paper aims to explain the process of personal selling management in an enterprise under the influence of current market conditions and global trends. The paper points out that there have been significant changes in the attitude towards work, management method, applying sophisticated tools, designing planning and strategic tasks, using information, etc. Today, a sales manager must possess the skill of making the right quality decisions, manage staff in an efficient and convincing way, have a sense of communication, clear the way for creativity and successfully solve problems.

A hybrid evolutionary algorithm integrated with specific neighborhood search for collaborative task planning of heterogeneous multi-UAVs under SEAD scenarios

A hybrid evolutionary algorithm integrated with specific neighborhood search for collaborative task planning of heterogeneous multi-UAVs under SEAD scenarios

Effects of collaborative vs. individual pre-task planning on EFL learners’ L2 writing: transferability of writing quality

Although many studies have investigated the differential effects of collaborative pre-task planning (CP), individual pre-task planning (IP), and no pre-task planning (NP) on EFL learners’ writing quality, none has explored such effects under the condition of combining IP with CP and whether such potential effects can be transferred to a new writing task. To bridge this gap, the study assigned 120 Chinese college learners to one of the four groups: CP group, IP group, IPCP group, and NP group. Two argumentative writing tasks were used in this study to collect the required data. To produce task 1, the learners in CP, IP, and IPCP groups were asked to do the relevant planning tasks before writing, while the learners in the NP group were required to write immediately. To explore learning transfer, one week after they produced task 1, the learners in all four groups were required to produce task 2, for which there was no planning time. The results showed that the learners under the CP condition composed writings with higher syntactic complexity than the learners under IP, IPCP, and NP conditions for both two tasks, with no such effects found for other measures of writing quality. Limitations, suggestions for future studies, and implications were discussed.

Multi-UAVs task allocation method based on MPSO-SA-DQN

Multi-UAVs play an important role in the battlefield. Although many methods are proposed to solve the Multi-UAV task allocation, there still existing the problems of complex time constraints and uncertain solution space. The reason is that multi-UAVs usually face changing environmental factors. Aiming at solving such problem, this paper proposes a multi-UAV task assignment method based on Deep Q-based evolutionary reinforcement learning algorithms (MPSO-SA-DQN). Specifically, this method builds a multi-agent training framework based on the deep evolutionary reinforcement learning mechanism and SA-DQN. Its aim is to improve the global exploration and optimization capabilities of multi-agents. At the same time, the multi-dimensional particle swarm optimization algorithm is introduced to optimize the state space. Based on task priority mapping, the MPSO-SA-DQN algorithm framework is proposed. As a result, multi-agents can optimize the execution state in real time in the environment interaction. Besides, it also has the ability to reach optimal state and maximum reward. According to the characteristics of multi-UAV global task assignment, this paper designs a priority state space autoencoder strategy and global task feature. A multi-UAVs tasks allocation and iterative optimization method based on MPSO-SA-DQN algorithm is proposed, so as to continuously optimize the task allocation scheme. The simulation results show that the multi-UAV task allocation method based on MPSO-SA-DQN can effectively solve the problem of uncertainty in the optimal solution space of task allocation. At the same time, the algorithm achieves faster convergence result, and a good prospect of promotion in the field of UAV swarm cooperative task planning.

Intelligent Swarm: Concept, Design and Validation of Self-Organized UAVs Based on Leader–Followers Paradigm for Autonomous Mission Planning

Unmanned Aerial Vehicles (UAVs), commonly known as drones, are omnipresent and have grown in popularity due to their wide potential use in many civilian sectors. Equipped with sophisticated sensors and communication devices, drones can potentially form a multi-UAV system, also called an autonomous swarm, in which UAVs work together with little or no operator control. According to the complexity of the mission and coverage area, swarm operations require important considerations regarding the intelligence and self-organization of the UAVs. Factors including the types of drones, the communication protocol and architecture, task planning, consensus control, and many other swarm mobility considerations must be investigated. While several papers highlight the use cases for UAV swarms, there is a lack of research that addresses in depth the challenges posed by deploying an intelligent UAV swarm. Against this backdrop, we propose a computation framework of a self-organized swarm for autonomous and collaborative missions. The proposed approach is based on the Leader–Followers paradigm, which involves the distribution of ROS nodes among follower UAVs, while leaders perform supervision. Additionally, we have integrated background services that autonomously manage the complexities relating to task coordination, control policy, and failure management. In comparison with several research efforts, the proposed multi-UAV system is more autonomous and resilient since it can recover swiftly from system failure. It is also reliable and has been deployed on real UAVs for outdoor survey missions. This validates the applicability of the theoretical underpinnings of the proposed swarming concept. Experimental tests carried out as part of an area coverage mission with 6 quadcopters (2 leaders and 4 followers) reveal that the proposed swarming concept is very promising and inspiring for aerial vehicle technology. Compared with the conventional planning approach, the results are highly satisfactory, highlighting a significant gain in terms of flight time, and enabling missions to be achieved rapidly while optimizing energy consumption. This gives the advantage of exploring large areas without having to make frequent downtime to recharge and/or charge the batteries. This manuscript has the potential to be extremely useful for future research into the application of unmanned swarms for autonomous missions.

Designing an optimal task scheduling and VM placement in the cloud environment with multi-objective constraints using Hybrid Lemurs and Gannet Optimization Algorithm

ABSTRACT An efficient resource utilization method can greatly reduce expenses and unwanted resources. Typical cloud resource planning approaches lack support for the emerging paradigm regarding asset management speed and optimization. The use of cloud computing relies heavily on task planning and allocation of resources. The task scheduling issue is more crucial in arranging and allotting application jobs supplied by customers on Virtual Machines (VM) in a specific manner. The task planning issue needs to be specifically stated to increase scheduling efficiency. The task scheduling in the cloud environment model is developed using optimization techniques. This model intends to optimize both the task scheduling and VM placement over the cloud environment. In this model, a new hybrid-meta-heuristic optimization algorithm is developed named the Hybrid Lemurs-based Gannet Optimization Algorithm (HL-GOA). The multi-objective function is considered with constraints like cost, time, resource utilization, makespan, and throughput. The proposed model is further validated and compared against existing methodologies. The total time required for scheduling and VM placement is 30.23%, 6.25%, 11.76%, and 10.44% reduced than ESO, RSO, LO, and GOA with 2 VMs. The simulation outcomes revealed that the developed model effectively resolved the scheduling and VL placement issues.

Mutation‐Guided Metamorphic Testing of Optimality in AI Planning

ABSTRACTAutonomous systems such as space‐ or underwater‐exploration robots or elderly people assistance robots often include an artificial intelligence (AI) planner as a component. Starting from the initial state of a system, an AI planner automatically generates sequential plans to reach final states that satisfy user‐specified goals. Generating plans having a minimum number of intermediate steps or taking the least time to execute is usually strongly desired, as these plans exhibit minimal costs. Unfortunately, testing if an AI planner generates optimal plans is almost impossible because the expected cost of these plans is usually unknown. Based on mutation adequacy test suite selection, this article proposes a novel metamorphic testing framework for detecting the lack of optimality in AI planners. The general idea is to perform a systematic but non‐exhaustive state space exploration from the initial state and to select mutant‐adequate states to instantiate new planning tasks as follow‐up test cases. We then check a metamorphic relation between the automatically generated solutions of the AI planner for these new test cases and the cost of the initial plan. We implemented this metamorphic testing framework in a tool called MorphinPlan. Our experimental evaluation shows that MorphinPlan can detect non‐optimal behaviour in both mutated AI planners and off‐the‐shelf, configurable planners. It also shows that our proposed mutation adequacy test selection strategy outperforms three alternative test generation and selection strategies, including both random state selection and random walks through the state space in terms of mutation scores.