Mobile Tasks Research Articles

Energy and latency aware mobile task assignment for green cloudlets

Edge computing places cloudlets with high computational capabilities near mobile devices to reduce the latency and network congestion encountered in cloud server-based task offloading. However, many cloudlets are required in such an edge computing network, leading to a tremendous increase in carbon emissions of computing networks globally. This increase in carbon emission envisages the need to employ green energy resources to power these cloudlets. This need has led to the concept of Green Cloudlet Networks (GCNs). But GCNs must deal with the problem of the unpredictability of green energy available to them while optimizing the performance (in terms of latency) delivered to the mobile user. This paper proposes a novel task-assignment called Green Energy and Latency Aware Task Assignment (Ge-LATA) for GCNs to address this issue. The primary aim of Ge-LATA is to optimize the latency and the green energy consumed in processing the offloaded tasks from the mobile devices. In this GCN, the cloudlets are connected in a network to process the incoming tasks cooperatively to ensure load-balancing at the cloudlets. Ge-LATA considers various factors like the current load, available green energy, service rate offered by cloudlets, and the distance from the mobile user, leading to optimal decisions in terms of latency and green energy consumed. Simulations are performed using the actual solar insolation data taken from the NREL database. Ge-LATA is tested with other offloading schemes for latency in processing the offloaded tasks and green-energy consumed under different solar insolation scenarios in these simulations. Simulation results show that Ge-LATA achieves up to 31.87% of reduction in the latency while ensuring up to 50.15% of reduction in the energy consumption than other comparable task-assignment schemes.Thus, Ge-LATA suggests that it leads to an optimal task assignment by considering the various factors mentioned above during the task assignment process. Thus, Ge-LATA considers the above-mentioned extensive set of parameters during the task allotment process. It also proposes an efficient green energy allotment scheme that adapts itself to actual weather and network conditions, leading to optimal task assignment decisions in GCNs.

DeepNet-Based 3D Visual Servoing Robotic Manipulation

The fourth industrial revolution (industry 4.0) demands high-autonomy and intelligence robotic manipulators. The goal is to accomplish autonomous manipulation tasks without human interventions. However, visual pose estimation of target object in 3D space is one of the critical challenges for robot-object interaction. Incorporating the estimated pose into an autonomous manipulation control scheme is another challenge. In this paper, a deep-ConvNet algorithm is developed for object pose estimation. Then, it is integrated into a 3D visual servoing to achieve a long-range mobile manipulation task using a single camera setup. The proposed system integrates (1) deep-ConvNet training using only synthetic single images, (2) 6DOF object pose estimation as sensing feedback, and (3) autonomous long-range mobile manipulation control. The developed system consists of two main steps. First, a perception network trains on synthetic datasets and then efficiently generalizes to real-life environment without postrefinements. Second, the execution step takes the estimated pose to generate continuous translational and orientational joint velocities. The proposed system has been experimentally verified and discussed using the Husky mobile base and 6DOF UR5 manipulator. Experimental findings from simulations and real-world settings showed the efficiency of using synthetic datasets in mobile manipulation task.

Zapline-plus: A Zapline extension for automatic and adaptive removal of frequency-specific noise artifacts in M/EEG.

Removing power line noise and other frequency‐specific artifacts from electrophysiological data without affecting neural signals remains a challenging task. Recently, an approach was introduced that combines spectral and spatial filtering to effectively remove line noise: Zapline. This algorithm, however, requires manual selection of the noise frequency and the number of spatial components to remove during spatial filtering. Moreover, it assumes that noise frequency and spatial topography are stable over time, which is often not warranted. To overcome these issues, we introduce Zapline‐plus, which allows adaptive and automatic removal of frequency‐specific noise artifacts from M/electroencephalography (EEG) and LFP data. To achieve this, our extension first segments the data into periods (chunks) in which the noise is spatially stable. Then, for each chunk, it searches for peaks in the power spectrum, and finally applies Zapline. The exact noise frequency around the found target frequency is also determined separately for every chunk to allow fluctuations of the peak noise frequency over time. The number of to‐be‐removed components by Zapline is automatically determined using an outlier detection algorithm. Finally, the frequency spectrum after cleaning is analyzed for suboptimal cleaning, and parameters are adapted accordingly if necessary before re‐running the process. The software creates a detailed plot for monitoring the cleaning. We highlight the efficacy of the different features of our algorithm by applying it to four openly available data sets, two EEG sets containing both stationary and mobile task conditions, and two magnetoencephalography sets containing strong line noise.

Association Between Socioeconomic Disadvantage and Decline in Function, Cognition, and Mental Health After Critical Illness Among Older Adults : A Cohort Study.

Older adults admitted to an intensive care unit (ICU) are at risk for developing impairments in function, cognition, and mental health. It is not known whether socioeconomically disadvantaged older persons are at greater risk for these impairments than their less vulnerable counterparts. To evaluate the association between socioeconomic disadvantage and decline in function, cognition, and mental health among older survivors of an ICU hospitalization. Retrospective analysis of a longitudinal cohort study. Community-dwelling older adults in the National Health and Aging Trends Study (NHATS). Participants with ICU hospitalizations between 2011 and 2017. Socioeconomic disadvantage was assessed as dual-eligible Medicare-Medicaid status. The outcome of function was defined as the count of disabilities in 7 activities of daily living and mobility tasks, the cognitive outcome as the transition from no or possible to probable dementia, and the mental health outcome as the Patient Health Questionnaire-4 score in the NHATS interview after ICU hospitalization. The analytic sample included 641 ICU hospitalizations for function, 458 for cognition, and 519 for mental health. After accounting for sociodemographic and clinical characteristics, dual eligibility was associated with a 28% increase in disability after ICU hospitalization (incidence rate ratio, 1.28; 95% CI, 1.00 to 1.64); and nearly 10-fold greater odds of transitioning to probable dementia (odds ratio, 9.79; 95% CI, 3.46 to 27.65). Dual eligibility was not associated with symptoms of depression and anxiety after ICU hospitalization (incidence rate ratio, 1.33; 95% CI, 0.99 to 1.79). Administrative data, variability in timing of baseline and outcome assessments, proxy selection. Dual-eligible older persons are at greater risk for decline in function and cognition after an ICU hospitalization than their more advantaged counterparts. This finding highlights the need to prioritize low-income seniors in rehabilitation and recovery efforts after critical illness and warrants investigation into factors leading to this disparity. National Institute on Aging.

Acute effects of axial loading on postural control during walking and turning in people with multiple sclerosis: A pilot study

BackgroundImpaired sensory integration is heavily involved in gait control and accentuates fall risk in individuals with multiple sclerosis (MS). While axial loading has been found beneficial, little is known about the effect of non-specific axial loads on gait parameters and mobility tasks in those with MS. Research Question: What are the effects of non-specific axial loading via weighted vests on walking and turning in those with MS. MethodsTwelve participants with MS and eleven age- and gender-matched healthy controls participated in a cross-sectional study. All participants completed five trials of continuous walking with turns wearing weighted vests at 0%, 2%, 4%, 5%, and then 0% of their body weight. Gait parameters were measured using wireless inertial sensors. A 2 (group) x 5 (vest weight) multivariate analysis of variance (MANOVA) was performed to determine any significant differences between groups and across weighted vests for each gait variable. Post-hoc analysis and paired t-tests with corresponding effect sizes were also conducted. ResultsA significant between groups main effect was found for group (F (6100) = 14.74, p = .000) in multiple gait parameters (p < 0.05), although no significant main effect was found for weighted vest. Within group analyses indicated significantly increased cadence and gait speed across varying weighted vests for both MS and control groups (p < 0 >05). Increased vest weight from 0%PRE to 2% also had large effect on shortening double support time and increasing stride length in the MS group. SignificanceThis study provided preliminary evidence that non-specific axial loads of varying weights appear to improve certain gait parameters. As such, this modality may offer mobility benefit and serve as an accessible home-based intervention alternative aimed at improving walking in individuals with MS.

Athlete-Perceived Impact of Frame Running on Physical Fitness, Functional Mobility and Psychosocial Outcomes.

ObjectiveFrame Running (RaceRunning) allows people with moderate-to-severe mobility impairments to participate in physical activity using a 3-wheeled frame with a saddle and handlebars. The aim of this study was to investigate athlete-perceived impact of Frame Running on aspects of physical fitness, functional mobility and psychosocial outcomes.DesignSurvey.ParticipantsFrame Running athletes aged 5 years and over.MethodsA survey was distributed to athletes through their club or sports organization.ResultsThe survey was completed by 115 athletes (53 females). Median age was 17 years (range 5–62 years) and 64 (57%) used a wheelchair or walker for distances over 50 m. Many felt that Frame Running stretched their muscles (n = 93, 87%) and increased their self-confidence (n = 63, 93%). Four (4%) reported extreme fatigue or sore muscles after training (n = 17, 15%). Of the 110 athletes who had been participating in Frame Running for over 3 months, 46 (47%) reported being less out of breath during mobility tasks and 66 (66%) felt they had improved their functional mobility. However, 7 (7%) reported increased muscle tightness and 4 (4%) reported a Frame Running-related injury lasting more than 4 weeks.ConclusionFrame Running is a safe physical activity with athlete-perceived benefits on physical fitness, functional mobility and psychosocial outcomes.LAY ABSTRACTFrame Running (RaceRunning) allows people with moderate-to-severe walking difficulties to walk or run independently using a 3-wheeled frame with a saddle and handlebars. In this study 115 Frame Running athletes living in the UK, Sweden or the Netherlands and aged 5 years or over (or their parents) completed a survey about taking part in Frame Running. More than half used a wheelchair or walker for distances over 50 m. Most felt that Frame Running stretched their muscles and increased their self-confidence. Four athletes perceived extreme fatigue and 17 reported sore muscles after training. Approximately half of the athletes reported that they were less out of breath during mobility tasks (walking, standing) and two thirds reported better functional mobility since they had started Frame Running. Injuries lasting more than 4 weeks were reported by 4 athletes. We conclude that Frame Running is a safe activity with the potential to improve physical fitness, functional mobility, and self-confidence.

Improving Wearable-Based Activity Recognition Using Image Representations

Activity recognition based on inertial sensors is an essential task in mobile and ubiquitous computing. To date, the best performing approaches in this task are based on deep learning models. Although the performance of the approaches has been increasingly improving, a number of issues still remain. Specifically, in this paper we focus on the issue of the dependence of today’s state-of-the-art approaches to complex ad hoc deep learning convolutional neural networks (CNNs), recurrent neural networks (RNNs), or a combination of both, which require specialized knowledge and considerable effort for their construction and optimal tuning. To address this issue, in this paper we propose an approach that automatically transforms the inertial sensors time-series data into images that represent in pixel form patterns found over time, allowing even a simple CNN to outperform complex ad hoc deep learning models that combine RNNs and CNNs for activity recognition. We conducted an extensive evaluation considering seven benchmark datasets that are among the most relevant in activity recognition. Our results demonstrate that our approach is able to outperform the state of the art in all cases, based on image representations that are generated through a process that is easy to implement, modify, and extend further, without the need of developing complex deep learning models.

Association between fall risk and assessments of single-task and dual-task walking among community-dwelling individuals with chronic stroke: A prospective cohort study

BackgroundFalling and fall-related injuries are common among community-dwelling individuals with chronic stroke. Falls often occur during dual-task walking scenarios. Accurate fall prediction is critical for formulating effective fall prevention strategies. Research questionsCan dual-task walking tests and corresponding single-task tests predict falls among individuals with chronic stroke? Are dual-task walking tests involving visuospatial cognition more effective in predicting falls than those involving other cognitive domains? MethodsNinety-three individuals with stroke (age: 62.4 ± 6.7 years; stroke duration: 5.6 ± 4.5 years) participated in this prospective cohort study. Two mobility tasks (level-ground walking and obstacle-crossing) were performed with and without two cognitive tasks (auditory clock test and auditory Stroop test). Demographic information and clinical measures of depression, motor function, walking speed and balance were collected. Monthly telephone interviews were conducted to collect data on fall incidence, related circumstances and injuries incurred during a 12-month follow-up period. Multivariate logistic regression analysis was performed to identify predictive factors associated with future risk of falls. ResultsThirty-six participants (39%) reported one or more falls during the follow-up period. The regression model including reaction time during the auditory clock task performance while negotiating obstacles correctly classified the fall status of 80% of the participants (72% future fallers and 84% non-fallers). Performance did not differ between fallers and non-fallers on any other measures tested. SignificanceDual-task assessment combining an auditory clock task with an obstacle-crossing task has potential clinical utility for identifying future fall risk among people with chronic stroke.

Linear actuator–based knee exoskeleton for stand–sit–stand motions: a bond graph approach

People with knee disorders often find it difficult to perform common mobility tasks, such as stand–sit–stand motions. High knee torque is required to complete such transitions, as the chances of toppling increase during these motions. Most of the existing conventional approaches, such as wheelchairs and crutches, have failed to provide complete independence to the users. Conversely, contemporary systems like lower body exoskeletons which are bulky, complex, and expensive do not specifically target the knee joint instead of assisting other joints. Hence, there is a need to aid the knee joint using a robotic knee exoskeleton capable of accurately providing the desired knee torque. In the present work, to assist the user in performing the stand–sit–stand motions, an electromyography sensor-based four-bar knee exoskeleton actuated by a linear actuator is proposed. The modeling of the complete exoskeleton is developed using bond graph technique, as the components exist in different energy domains and it is possible to frame a dynamic bond graph model using only kinematic equations. The prototype is fabricated, and experiments are carried out on an artificial limb to prove the efficacy of the design of the current knee exoskeleton. The assistive torque developed by the actuator at the knee joint of the exoskeleton is found to be suitable to assist the wearer. As a result, little effort is required by the wearer for performing the stand–sit–stand motions. The rotation of the thigh link of the developed exoskeleton was found to be suitable for performing the stand–sit–stand activity.

Integrating a multi-disciplinary mobility program into a Long-Term Care Practice: A quality initiative to improve value

Background: The majority of patients living in a Long-Term Care Unit (LTCU) have limited mobility.1 Prolonged immobilization results in functional decline and increases the risk of other complications.2 It was observed that there is a lack of a structured, evidence-based mobility program using mobility protocols and guidelines. As per a 6-weeks baseline data collection, out of the 36 patients in Rumailah Hospital&apos;s LTCU-2, 40% were bedbound and only 60% of them were engaged in any kind of “Out Of Bed Therapeutic Activities” (OOBTA) as of November 2020.The project aimed to increase patients’ participation in OOBTA from 60% to 80% by December 2021. Methods: A Mobility Task Force (MTF) was formed from a Multi-disciplinary Team (MDT). The Institute for Healthcare Improvement (IHI) Model of Improvement was utilized. A Cause and Effect Diagram with multi-disciplinary input identified potential causes limiting mobility and supported possible solutions for testing in a series of Plan-Do-Study-Act (PDSA) cycles (Figure 1). Each specialty evaluated and provided appropriate OOBTA based on their level of care. Results: This project helped the unit to promote a culture3 of early and safe patient mobility as shown by increased participation in OOBTA and zero falls, no pressure injuries, etc. As a result, it helped the unit to attain more than 80% of patient participation in OOBTA during their hospital stay (Figure 2). The team observed good multidisciplinary collaboration and engagement in developing this mobility program. Conclusion: Promoting early mobility is a cultural change that improved patients’ participation in OOBTA from 60% to 80% from July 2021, five months ahead of our target date and despite the COVID-19 pandemic situation.

User Recruitment Algorithm for Maximizing Quality under Limited Budget in Mobile Crowdsensing

In the mobile crowdsensing task assignment, under the premise that the data platform does not know the user’s perceived quality or cost value, how to establish a suitable user recruitment mechanism is the critical issue that this article needs to solve. It is necessary to learn the user’s perceived quality in the execution p. It also needs to try its best to ensure the efficiency and profit maximization of the mobile group intelligence perception platform. Therefore, this paper proposes a mobile crowdsensing user recruitment algorithm based on Combinatorial Multiarmed Bandit (CMAB) to solve the recruitment problem with known and unknown user costs. Firstly, the user recruitment process is modeled as a combined multiarm bandit model. Each rocker arm represents the selection of different users, and the income obtained represents the user’s perceived quality. Secondly, it proposes the upper confidence bound (UCB) algorithm, which updates the user’s perceptual quality according to the completion of the task. This algorithm sorts the users’ perceived quality values from high to low, then selects the most significant ratio of perceived quality to recruitment costs under the budget condition, assigns tasks, and updates their perceived quality. Finally, this paper introduces the regret value to measure the efficiency of the user recruitment algorithm and conducts many experimental simulations based on real data sets to verify the feasibility and effectiveness of the algorithm. The experimental results show that the recruitment algorithm with known user cost is close to the optimal algorithm, and the recruitment algorithm with unknown user cost is more than 75% of the optimal algorithm result, and the gap tends to decrease as the budget cost increases, compared with other comparisons. The algorithm is 25% higher, which proves that the proposed algorithm has good learning ability and can independently select high‐quality users to realize task assignments.

Velocity clamping-assisted adaptive salp swarm algorithm: balance analysis and case studies.

Salp swarm algorithm (SSA) is a recently proposed, powerful swarm-intelligence based optimizer, which is inspired by the unique foraging style of salps in oceans. However, the original SSA suffers from some limitations including immature balance between exploitation and exploration operators, slow convergence and local optimal stagnation. To alleviate these deficiencies, a modified SSA (called VC-SSA) with velocity clamping strategy, reduction factor tactic, and adaptive weight mechanism is developed. Firstly, a novel velocity clamping mechanism is designed to boost the exploitation ability and the solution accuracy. Next, a reduction factor is arranged to bolster the exploration capability and accelerate the convergence speed. Finally, a novel position update equation is designed by injecting an inertia weight to catch a better balance between local and global search. 23 classical benchmark test problems, 30 complex optimization tasks from CEC 2017, and five engineering design problems are employed to authenticate the effectiveness of the developed VC-SSA. The experimental results of VC-SSA are compared with a series of cutting-edge metaheuristics. The comparisons reveal that VC-SSA provides better performance against the canonical SSA, SSA variants, and other well-established metaheuristic paradigms. In addition, VC-SSA is utilized to handle a mobile robot path planning task. The results show that VC-SSA can provide the best results compared to the competitors and it can serve as an auxiliary tool for mobile robot path planning.

Biomechanical Adaptations to Lower Limb Amputation during Functional Mobility Tasks

Biomechanical Adaptations to Lower Limb Amputation during Functional Mobility Tasks

THE IMPLEMENTATION OF MOBILE VIDEO TASK TECHNIQUE IN TEACHING SPEAKING AT THE ELEVENTH GRADE STUDENTS OF SMKN 1 KEMPO

The aimed of this research was to find out the improve of students’ speaking ability by using Mobile Video Task that focused on students’ fluency at the Eleventh Grade of SMKN 1 Kempo. The researcher used Pre-experimental Design with one group pre-test and post-test, and data collection based on the test. The sample of this research was class XI which consisted of 20 students. The sample was taken by using Purposive Sampling Technique. The research findings showed that the students’ difference before and after using mobile video task was significantly different. The students’ pre-test mean score in term of fluency was 65.15 to be 75.45 in post-test. Based on the findings, it can be concluded that mobile video task technique was effective to improve the speaking ability of the students at the Eleventh Grade of SMKN1 Kempo in the Academic Year 2020/2021.

Online Discrete Anchor Graph Hashing for Mobile Person Re-Identification

With the advance of mobile technologies, mobile devices such as unmanned aerial vehicle (UAV) become more important in video surveillance. By applying mobile person re-identification (re-id), mobile devices can monitor pedestrians in the transportation system from complex environments. Since the computing and storage resources of mobile devices are limited, traditional person re-id methods are not appropriate for mobile condition. Besides, mobile person re-id task also requires real-time processing. In this paper, we propose a novel hashing method: online discrete anchor graph hashing (ODAGH) for mobile person re-id. ODAGH integrates the advantages of online learning and hashing technology. In ODAGH, we propose an online discrete optimization algorithm to improve the efficiency of anchor graph learning in the online scenario. Experimental results demonstrate the superiority of ODAGH in terms of both effect and efficiency.

Actions Selection During a Mobile Robot Navigation for the Autonomous Recharging Problem

The use of mobile robots has increased for its application in various areas such as supply chains, factories, cleaning, disinfection, medical assistance, search, and exploration. It is a fact that most of these robots, if not all, use batteries to power themselves. During a mobile robot task execution, the battery's electric charge tends to deplete as a function of the energy load demands, which would cause the robot to shut down if the discharge is critical, leaving its task inconclusive. Therefore, it is of utmost importance that the robot learns when to charge its batteries, avoiding turning off. This work shows a reactive navigation scheme for a mobile robot that integrates a module for battery-level monitoring. A robot moves from a starting point to a destination according to the battery level. During the navigation, the robot decides when to change the course toward a battery charging station. This paper presents a rules-based reinforcement learning architecture with three entries; these entries correspond to the robot's battery level, the distance to the destination, and the distance to the battery charging station. According to the simulations, the robot learns to select an appropriate action to accomplish its task.

Motivating Selective Motor Control of Infants at High Risk of Cerebral Palsy Using an In-Home Kicking-Activated Mobile Task: A Pilot Study.

Decreased selective motor control limits gait function of children with spastic cerebral palsy (CP). Infants at high risk of CP demonstrate decreased selective motor control by 1month of age. To motivate more selective hip-knee control, infants at high risk of CP participated in an in-home kicking-activated mobile task. The purpose of this study was to determine whether infants at high risk of CP and infants with typical development (TD) demonstrated increased selective hip-knee control during 2-minute intervals of the mobile task when they demonstrated learning of the association between their leg movement and mobile activation vs during 2-minute intervals when they did not demonstrate learning. Participants in this cohort study included 10 infants at high risk of CP based on neuroimaging and 11 infants with TD at 3.5 to 4.5months of age. Each infant participated in the in-home kicking-activated mobile task for 8 to 10min/d, 5 d/wk, for 6weeks. Over 80,000 kicks were extracted and classified for each infant as occurring during 2-minute intervals of the task when the infant demonstrated learning vs not learning based on mobile activation time above baseline. Infants demonstrated kicks with more selective hip-knee control during 2-minute intervals of the mobile task when they demonstrated learning compared with when they did not demonstrate learning for 4 of 6weeks in the cohort at high risk of CP and for 2 of 6weeks in the cohort with TD. Participation in the in-home kicking-activated mobile task may motivate more selective hip-knee control of infants at high risk of CP. This study is a first step toward developing an intervention to promote selective hip-knee control of infants at high risk of CP, with the ultimate goal of optimizing future walking function. This study showed that playing with an in-home infant kicking-activated mobile may motivate infants at high risk of CP to produce more age-appropriate leg movements.

The Effect Of Mobile Phone Use While Driving On Response Time: Driving Simulator Study

Mobile phone use is one of the most common daily tasks and this is normal, however, this task could be problematic while driving. The use of mobile phones while driving has become a major cause of road accidents and poses a threat to public health. This study investigated the effect of mobile phone usage while driving on response time, as it investigated four mobile phone tasks (hands-free calling, hand calls, reading text messages, and sending text messages) in addition to basic driving. A total of 42 participants, ranging in age from (19 to 55), with a mean age (mean = 33.14, SD = 10.26) participated in the driving simulation at the University of Technology and all participants performed five tasks. The participants had to interact with voice commands by performing the throttle maneuver. The results concluded with a delay in response, which means an increase in cognitive reaction time when using a mobile phone compared to basic driving. It has also been found that the response time increases with the age of drivers.

Game Combined Multi-Agent Reinforcement Learning Approach for UAV Assisted Offloading

Air ground integrated mobile cloud computing (MCC) provides unmanned aerial vehicles (UAVs) the capability to act as an aerial relay with more flexibility and resilience. In the cloud computing architecture, the data generated by ground users (GUs) can be offloaded to the remote server for fast processing. However, the heterogeneity of mobile tasks makes the data size distributed among GUs unbalanced. Besides, the energy efficiency of UAVs movement should be carefully considered for sustainable flight and obstacle avoidance. In general, such a joint trajectory issue can hardly be formulated as a convex optimization in unpredictable and dynamic environments. This paper proposes a potential game combined multi-agent deep deterministic policy gradient (MADDPG) approach to optimize multiple UAVs' trajectory with the consideration of GUs' offloading delay, energy efficiency as well as obstacle avoidance system. In specific, we first model the issue as a mixed integer non-linear problem (MINP), in which the service assignment between multi-user and multi-UAV is solved by potential game. The convergence to a Nash Equilibrium (NE) can be achieved by distributive service assignment update with infinite iteration. Then, we optimize the trajectory with obstacle avoidance at each UAV by MADDPG approach, which has a great advantage of centralized-training and decentralized-execution to reduce the global synchronized communication overhead. UAVs movement can be optimized in continuity rather than other deep reinforcement learning (DRL) approaches generating discrete simple actions. Experiments demonstrate the proposed game-combined learning algorithm can minimize the offloading delay, enhance UAVs’ energy efficiency and avoid the obstacles at the same time.

Unsupervised deep learning based ego motion estimation with a downward facing camera

Knowing the robot's pose is a crucial prerequisite for mobile robot tasks such as collision avoidance or autonomous navigation. Using powerful predictive models to estimate transformations for visual odometry via downward facing cameras is an understudied area of research. This work proposes a novel approach based on deep learning for estimating ego motion with a downward looking camera. The network can be trained completely unsupervised and is not restricted to a specific motion model. We propose two neural network architectures based on the Early Fusion and Slow Fusion design principle: “EarlyBird” and “SlowBird”. Both networks share a Spatial Transformer layer for image warping and are trained with a modified structural similarity index (SSIM) loss function. Experiments carried out in simulation and for a real world differential drive robot show similar and partially better results of our proposed deep learning based approaches compared to a state-of-the-art method based on fast Fourier transformation.