Collision Avoidance Research Articles

Safedrive dreamer: Navigating safety–critical scenarios in autonomous driving with world models

Achieving stable and reliable autonomous driving in complex traffic environments while ensuring safety under unpredictable conditions is a critical challenge in autonomous driving technology. To address this issue, this study proposes the Safedrive Dreamer navigation framework, which aims to reduce the reliance on trial-and-error learning in real-world scenarios, thereby mitigating the risks associated with dynamic driving conditions and enhancing vehicle foresight. This framework integrates the predictive capabilities of world models with the constrained Markov decision process (CMDP) and safety reinforcement learning to accurately anticipate future environmental changes. This ensures the reliability of autonomous driving routes, thereby improving both safety and efficiency. Furthermore, to reduce trial-and-error costs in real-world applications, this study employs PAC-Bayesian methods to derive generalization error bounds between simulations and reality, enabling a more effective transfer of knowledge and experience from simulations to real-world scenarios. Validation experiments in simulated and real environments showed that Safedrive Dreamer significantly outperformed existing autonomous driving solutions by 3.8% in key safety metrics, excelling in collision avoidance and risk reduction. This study provides new insights into the integration of world models into decision-making processes to enhance decision-making capabilities in safety–critical applications, thereby contributing significantly to the improvement of autonomous driving safety and reliability.

Path Planning of an Unmanned Aerial Vehicle Based on a Multi-Strategy Improved Pelican Optimization Algorithm.

The UAV path planning algorithm has many applications in urban environments, where an effective algorithm can enhance the efficiency of UAV tasks. The main concept of UAV path planning is to find the optimal flight path while avoiding collisions. This paper transforms the path planning problem into a multi-constraint optimization problem by considering three costs: path length, turning angle, and collision avoidance. A multi-strategy improved POA algorithm (IPOA) is proposed to address this. Specifically, by incorporating the iterative chaotic mapping method with refracted reverse learning strategy, nonlinear inertia weight factors, the Levy flight mechanism, and adaptive t-distribution variation, the convergence accuracy and speed of the POA algorithm are enhanced. In the CEC2022 test functions, IPOA outperformed other algorithms in 69.4% of cases. In the real map simulation experiment, compared to POA, the path length, turning angle, distance to obstacles, and flight time improved by 8.44%, 5.82%, 4.07%, and 9.36%, respectively. Similarly, compared to MPOA, the improvements were 4.09%, 0.76%, 1.85%, and 4.21%, respectively.

Research and innovation of the “ship navigation radar” course based on the four‐in‐one approach in engineering education accreditation

AbstractTo achieve the strategic goal of marine and transportation power, it is imperative to cultivate highly skilled navigation professionals with a strong theoretical foundation and practical expertize. Seizing the opportunity by the “Double Thousand Plan” first‐class undergraduate courses initiated by the Ministry of Education, a four‐in‐one experimental teaching model is developed for ship navigation radar courses that integrate virtual reality technology under the context of engineering certification. The model encompasses land laboratory real training, maritime radar simulator training platform, comprehensive real‐ship training at sea, and virtual laboratory teaching and training. Through this innovative approach, students can acquire essential knowledge of ship navigation radar required for practical activities in navigation practice while honing their operational skills in positioning, navigation, and collision avoidance using radars. Furthermore, an analysis of course attainment and future development trends is conducted to provide insights into enhancing maritime education and fostering internationally competitive shipping talents.

Navigation of a Quadrotor Based on Voronoi Division Calculated from Local Information

This study applies a successful collision-avoidance method using buffered Voronoi cells (BVC) to control quadrotors. In particular, we consider the case of dealing with stationary obstacles using only the local information obtained from sensors, which has not been discussed in previous studies. In this study, we assume an unknown environment with grid-shaped obstacles. We demonstrate that four quadrotors can compute the information required for the BVC-based collision avoidance algorithm and move in the same environment without communicating with each other or receiving information from a central controller, using the local information obtained from the mounted 3D LiDAR. Simulations indicate that the system can handle static obstacles using point-cloud data obtained using 3D LiDAR as Voronoi seeds. We also demonstrate that the BVC-based collision avoidance algorithm can be applied to quadrotors controlled by a simple PD position controller without any modification to the controller. These findings show that the BVC-based collision avoidance can be easily implemented using existing commercial drones.

The crowd cooperation approach for formation maintenance and collision avoidance using multi-agent deep reinforcement learning

The crowd cooperation approach for formation maintenance and collision avoidance using multi-agent deep reinforcement learning

Mathematical Model of Horizontal Track Conflict Prevention Algorithm in Detect-and-Avoid Framework

With the proliferation of Unmanned Aerial Vehicle (UAV) technology, the demand for effective collision avoidance technology has intensified. The DAIDALUS algorithm, devised by NASA Langley Research Center under the Detect-and-Avoid (DAA) framework, provides conflict prevention bands for remotely piloted UAVs navigating in intricate airspace. The algorithm computes the bands in two distinct phases: Conflict and Recovery. The formal model for both phases has been established and implemented through iterative programming approaches. However, the mathematical model remains incomplete. Therefore, based on the model, this paper proposes the mathematical model for the two phases of the horizontal track conflict prevention algorithm. Firstly, Cauchy’s inequality is proposed to formulate the model that addresses trajectory conflicts considering the UAV non-instantaneous maneuvering dynamics model, and then a prudent maneuvering strategy is designed to optimize the model for the recovery phase. Finally, the execution procedure of the algorithm within the two-stage mathematical model is also detailed. The results demonstrate that the proposed model achieves a higher precision in the preventive bands, implements an effective collision avoidance strategy, and consistently aligns with the DAIDALUS model while offering a larger buffer time or distance. This work theoretically validates the formal model of the DAIDLAUS algorithm and provides insights for further refinement.

Digital Traffic Lights: UAS Collision Avoidance Strategy for Advanced Air Mobility Services

With the advancing development of Advanced Air Mobility (AAM), there is a collaborative effort to increase safety in the airspace. AAM is an advancing field of aviation that aims to contribute to the safe transportation of goods and people using aerial vehicles. When aerial vehicles are operating in high-density locations such as urban areas, it can become crucial to incorporate collision avoidance systems. Currently, there are available pilot advisory systems such as Traffic Collision and Avoidance Systems (TCAS) providing assistance to manned aircraft, although there are currently no collision avoidance systems for autonomous flights. Standards Organizations such as the Institute of Electrical and Electronics Engineers (IEEE), Radio Technical Commission for Aeronautics (RTCA), and General Aviation Manufacturers Association (GAMA) are working to develop cooperative autonomous flights using UAS-to-UAS Communication in structured and unstructured airspaces. This paper presents a new approach for collision avoidance strategies within structured airspace known as “digital traffic lights”. The digital traffic lights are deployed over an area of land, controlling all UAVs that enter a potential collision zone and providing specific directions to mitigate a collision in the airspace. This strategy is proven through the results demonstrated through simulation in a Cesium Environment. With the deployment of the system, collision avoidance can be achieved for autonomous flights in all airspaces.

Unmanned Ship Collision Avoidance Action Plan Deduction Method under Man–Machine Interactive Negotiation in Collision Avoidance Scenarios

With the development of artificial intelligence technology, the future water traffic environment will present a new pattern of coexistence of manned ships and unmanned ships, because unmanned ships are different from manned ships in situation understanding, collision avoidance decision-making, and so on. Therefore, the obstacle avoidance planning between unmanned ships and manned ships becomes extremely complex, and collision avoidance behavior scheme deduction becomes a key step in solving the problems related to situation understanding and collision avoidance decision-making in collision avoidance scenarios. In this paper, the situation understanding of the pilot for different collision avoidance situations is integrated into the dynamic obstacle avoidance model, and an intelligent navigation collision avoidance system is proposed to assist in deducing the collision avoidance action plan of the unmanned ship in the man–machine coexistence scenario. The intelligent navigation collision avoidance system is divided into two parts, namely a ship situation understanding part and a ship obstacle avoidance part, wherein ship situation understanding is used for realizing the transition of the collision state of the unmanned ship in the deduction process by constructing a collision-state set and a behavior decision set by using a finite state machine (FSM). Regarding ship obstacle avoidance, ship velocity obstacle is calculated based on the reciprocal velocity obstacle method (RVO), and the collision avoidance action is selected by using the behavior decision generated by the FSM to realize the dynamic collision avoidance deduction of the unmanned ship. In this study, the validity and effectiveness of the intelligent navigation collision avoidance system proposed in this paper are verified by case studies in a variety of collision avoidance scenarios. The system successfully solves the problem of intelligent collision avoidance planning, provides reliable support for the intelligent collision avoidance of unmanned ships, provides a feasible solution for safety and efficiency in sea navigation, and provides a valuable reference for the design and development of future intelligent navigation collision avoidance systems for ships.

A novel collaborative collision avoidance decision method for multi-ship encounters in complex waterways

With the continuous growth of maritime traffic volume, multi-ship collision avoidance has gradually become one of the most crucial issues in ensuring the safety of ship navigation. This study proposes a multi-ship collaborative collision avoidance decision model (MCCA) to address the issue of multi-ship encounters in complex waterways. The model takes into account the ship spatio-temporal collision risk, relative motion trends, as well as the safety and economic considerations of ship navigation. A comprehensive collision avoidance method is designed based on the velocity obstacle method, enabling simultaneous adjustments both course and speed, with the aim of minimizing economic losses incurred by ships due to avoidance operations while ensuring ship navigation safety. Meanwhile, the NSGA-II algorithm, which has faster convergence and better effect, is used to calculate the optimal collaborative collision avoidance decision of multiple ships. Besides, both simulation and example validation are used to test the validity of the model. The results indicate that the proposed model can effectively achieve multi-ship collaborative collision avoidance in the region. Simultaneously, it holds significant reference value for guaranteeing the safety of ship navigation, enhancing supervision efficiency, and promoting the development of intelligent ship navigation systems.

Use of neural networks for safe collision avoidance for groups of autonomous vessels

This article analyzes and compares the effectiveness of two algorithms for collision avoidance in groups of autonomous vessels: one based on geometric analysis of vessel approach and cost function minimization for calculating a safe maneuver (traditional algorithm), and an algorithm utilizing a neural network. Both algorithms assume external control of a group of vessels within a specific area using a vessel traffic management system and coordinated maneuvering of dangerously approaching vessels. Descriptions of these algorithms are provided along with their simplified block diagrams; to solve the task of safely maneuvering a group of vessels, a sequential analysis of all possible vessel pairs in the group and changes in their courses is proposed. The process of creating three test datasets is described, two of which were generated using a program and included 100 scenarios each, while the third was manually composed and included 30 scenarios for different vessel group approach variations. During the testing of the neural network algorithm, two neural networks trained to predict safe courses for vessel pairs were utilized. The neural network used in the algorithm, trained on 743671 samples, allowed the processing of test vessel approach scenarios with an accuracy comparable to the traditional algorithm. Depending on the number of dangerously approaching vessels in the area, the neural network algorithm processed test scenarios 2–14 times faster than the traditional algorithm. The paper highlights the limitations of the described algorithms and outlines planned improvements for subsequent research, including the optimization of the safe maneuver selection methodology and further training of the neural network on larger data volumes.

A Fuzzy Fusion Method for Multi-Ship Collision Avoidance Decision-Making with Merchant and Fishing Vessels

In multi-vessel collision avoidance decision-making, the collision between merchant and fishing vessels is a significant challenge. This paper proposes a fuzzy fusion method for making avoidance decisions under the influence of the navigation environment. First, C-means clustering was used to collect and analyze Automatic Identification System (AIS) data from fishing vessels. On this basis, the environment collision risk was determined using fuzzy reasoning. Second, the basic collision risk is obtained by calculating the DCPA and TCPA, and the integrated Collision Risk Index (CRI) is concluded by fuzzy logic through basic collision risk and the environment collision risk. The similar cases are extracted from the fuzzy case database, and collision avoidance decisions for merchant vessels are formulated following fuzzy adjustments. Finally, to validate the method, data from Chengshantou coastal waters is employed for verification. The results show that it can provide theoretical guidance and practical value for merchant vessels in making collision avoidance decisions.

Energy-Aware Hierarchical Control of Joint Velocities

Nowadays, robots are applied in dynamic environments. For a robust operation, the motion planning module must consider other tasks besides reaching a specified pose: (self) collision avoidance, joint limit avoidance, keeping an advantageous configuration, etc. Each task demands different joint control commands, which may counteract each other. We present a hierarchical control that, depending on the robot and environment state, determines online a suitable priority among those tasks. Thereby, the control command of a lower-prioritized task never hinders the control command of a higher-prioritized task. We ensure smooth control signals also during priority rearrangement. Our hierarchical control computes reference joint velocities. However, the underlying concepts of hierarchical control differ when using joint accelerations or joint torques as control signals instead. So, as a further contribution, we provide a comprehensive discussion on how joint velocity control, joint acceleration control, and joint torque control differ in hierarchical task control. We validate our formulation in an experiment on hardware.

UAV p­­­arcel Delivery System with Deep Reinforcement Learning Based Collision Avoidance and Route Optimization

Unmanned Aerial Vehicles (UAVs), or drones, have recently become a favorable solution for fast parcel delivery due to their maneuverability and advances in navigation technologies. With the limitation of battery capacity and payload of drones, it is crucial to consider both efficiency and cost while conducting the tasks. Meanwhile, UAVs should not collide with each other while traveling to customers. In this article, we propose a UAV parcel delivery system involving deep reinforcement learning (DRL) approach for collision avoidance and a genetic algorithm for route optimization. Specifically, a delivery center generates near-optimal routes, loading UAV with parcels according to demands. Each UAV takes charge of delivering packages in compliance with the assigned route while avoiding collision with each other. We utilize DRL to achieve collision avoidance without having prior knowledge about the trajectories of other UAVs. Additionally, we adopt a genetic algorithm to obtain the lowest energy cost path for each UAV. To find such an optimized path, we solve a capacitated vehicle routing problem (CVRP) with a modified cost function and extra constraints. Realistic simulations using a physics engine and software-in-the-loop (SITL) are conducted to evaluate the feasibility of the proposed methods.

A comprehensive survey of space robotic manipulators for on-orbit servicing.

On-Orbit Servicing (OOS) robots are transforming space exploration by enabling vital maintenance and repair of spacecraft directly in space. However, achieving precise and safe manipulation in microgravity necessitates overcoming significant challenges. This survey delves into four crucial areas essential for successful OOS manipulation: object state estimation, motion planning, and feedback control. Techniques from traditional vision to advanced X-ray and neural network methods are explored for object state estimation. Strategies for fuel-optimized trajectories, docking maneuvers, and collision avoidance are examined in motion planning. The survey also explores control methods for various scenarios, including cooperative manipulation and handling uncertainties, in feedback control. Additionally, this survey examines how Machine learning techniques can further propel OOS robots towards more complex and delicate tasks in space.

Pose-Constrained Control of Proximity Maneuvering for Tracking and Observing Noncooperative Targets with Unknown Acceleration

This paper proposes a pose control scheme of for proximity maneuvering for tracking and observing noncooperative targets with unknown acceleration, which is an important prerequisite for on-orbit operations in space. It mainly consists of a finite-time extended state observer and constraint processing procedures. Firstly, relative pose-coupled kinematics and dynamics models with unknown integrated disturbances are established based on dual quaternion representations. Then, a finite-time extended state observer is designed using the super-twisting algorithm to estimate the integrated disturbances. Both observation field of view and collision avoidance pose-constrained models are constructed to ensure that the service spacecraft continuously and safely observes the target during proximity maneuvering. And the constraint models are further incorporated into the design of artificial potential function with a unique minimum. After that, the proportional–derivative-like pose-constrained tracking control law is proposed based on the estimated disturbances and the gradient of the artificial potential function. Finally, the effectiveness of the control scheme is verified through numerical simulations.

A Risk Identification Method for Ensuring AI-Integrated System Safety for Remotely Controlled Ships with Onboard Seafarers

The maritime sector is increasingly integrating Information and Communication Technology (ICT) and Artificial Intelligence (AI) technologies to enhance safety, environmental protection, and operational efficiency. With the introduction of the MASS Code by the International Maritime Organization (IMO), which regulates Maritime Autonomous Surface Ships (MASS), ensuring the safety of AI-integrated systems on these vessels has become critical. To achieve safe navigation, it is essential to identify potential risks during the system planning stage and design systems that can effectively address these risks. This paper proposes RA4MAIS (Risk Assessment for Maritime Artificial Intelligence Safety), a risk identification method specifically useful for developing AI-integrated maritime systems. RA4MAIS employs a systematic approach to uncover potential risks by considering internal system failures, human interactions, environmental conditions, AI-specific characteristics, and data quality issues. The method provides structured guidance to identify unknown risk situations and supports the development of safety requirements that guide system design and implementation. A case study on an Electronic Chart Display and Information System (ECDIS) with an AI-integrated collision avoidance function demonstrates the applicability of RA4MAIS, highlighting its effectiveness in identifying specific risks related to AI performance and reliability. The proposed method offers a foundational step towards enhancing the safety of software systems, contributing to the safe operation of autonomous ships.

Optimized Right-Turn Pedestrian Collision Avoidance System Using Intersection LiDAR

The incidence of right-turning pedestrian accidents is increasing in South Korea. Most of the accidents occur when a large vehicle is turning right, and the main cause of the accidents was found to be the driver’s limited field of vision. After these accidents, the government implemented a series of institutional measures with the objective of preventing such accidents. However, despite the institutional arrangements in place, pedestrian accidents continue to occur. We focused on the many limitations that autonomous vehicles, like humans, can face in such situations. To address this issue, we propose a right-turn pedestrian collision avoidance system by installing a LiDAR sensor in the center of the intersection to facilitate pedestrian detection. Furthermore, the urban road environment is considered, as this provides the optimal conditions for the model to perform at its best. During this research, we collected data on right-turn accidents using the CARLA simulator and ROS interface and demonstrated the effectiveness of our approach in preventing such incidents. Our results suggest that the implementation of this method can effectively reduce the incidence of right-turn accidents in autonomous vehicles.

Synthesis of dynamic modelling framework and optimal control strategy for virtually coupled train sets with guaranteed safety

Virtual coupling represents a novel railway operational mechanism, wherein several trains share state information and adjust their behaviours based on control objectives, which include safety (collision avoidance) assurance and stability (uniform velocity and constant spacing between trains) maintenance. However, the conflict between safety and stability requirements becomes more apparent, as the closer spacing between trains. Moreover, the hybrid nature of the train platoon operation, which involves both discrete events and continuous dynamics, also poses a significant barrier to the control strategy synthesis. This paper presents the Virtually Coupled Train Sets (VCTS) dynamic modelling framework and strategy synthesis approach to address these problems. A Hybrid Markov Decision Process (HMDP) is adopted for relative longitudinal dynamics modelling of trains, considering the influence of the uncertainty dynamic of trains ahead. A Stackelberg game-based strategy synthesis algorithm is applied for the safe guarantee by overcoming the uncertainty, and a Q-learning method is used for stability strategy selection from the safe strategy. The results showed a 17.95% expansion in the safe state space when compared to the general relative braking scheme (which assumes maximal acceleration during the delay horizon). In a four-train tracking operation scenario, compared with Q-learning without safe constraints, our approach gives similar stability performance; compared with Q-learning with general relative braking safe constraints, our approach not only assurance safety but has a better stability performance.

Ego-Vehicle Speed Correction for Automotive Radar Systems Using Convolutional Neural Networks.

The development of autonomous driving vehicles has increased the global demand for robust and efficient automotive radar systems. This study proposes an automotive radar-based ego-vehicle speed detection network (AVSD Net) model using convolutional neural networks for estimating the speed of the ego vehicle. The preprocessing and postprocessing methods used for vehicle speed correction are presented in detail. The AVSD Net model exhibits characteristics that are independent of the angular performance of the radar system and its mounting angle on the vehicle, thereby reducing the loss of the maximum detection range without requiring a downward or wide beam for the elevation angle. The ego-vehicle speed is effectively estimated when the range-velocity spectrum data are input into the model. Moreover, preprocessing and postprocessing facilitate an accurate correction of the ego-vehicle speed while reducing the complexity of the model, enabling its application to embedded systems. The proposed ego-vehicle speed correction method can improve safety in various applications, such as autonomous emergency braking systems, forward collision avoidance assist, adaptive cruise control, rear cross-traffic alert, and blind spot detection systems.

Research on Autonomous Collision Avoidance of USV Based on Improved APF Algorithm

Abstract Aiming at the problem of Unmanned Surface Vessel (USV) avoiding dynamic vessels on the sea surface, an Artificial Potential Field method based on Velocity Obstacle Method (VO-APF) was studied. Under the premise of complying with the Convention on the International Regulations for Preventing Collisions at Sea (COLREGs), the overlapping of ship domains is minimized. The Quaternion ship domain was introduced into VO to make it more suitable for collision avoidance judgment in the ship domain, and the collision avoidance time of the ship was predicted, and the reasonable obstacle avoidance point was finally obtained. The attraction function of the obstacle avoidance target point and the repulsion function of the ship domain based on the APF are set, and the USV is subjected to the repulsion force and attraction in the APF, and finally completes the whole dynamic ship collision avoidance task. In the simulation environment, VO, APF and VO-APF were compared in three different encounter situations. The results show that VO-APF can help USV to complete the collision avoidance task under the premise of complying with COLREGs, and the overlap time between USV’s own ship domain and the other ship domain is the shortest, and the safety degree is the highest.