Learning from demonstration for locally assistive mobility aids

Abstract

Active assistive systems for mobility aids are largely restricted to environments mapped a-priori, while passive assistance primarily provides collision mitigation and other hand-crafted behaviors in the platform's immediate space. This paper presents a framework providing active short-term assistance, combining the freedom of location independence with the intelligence of active assistance. Demonstration data consisting of on-board sensor data and driving inputs is gathered from an able-bodied expert maneuvring the mobility aid around a generic interior setting, and used in constructing a probabilistic intention model built with Radial Basis Function Networks. This allows for short-term intention prediction relying only upon immediately available user input and on-board sensor data, to be coupled with real-time path generation based upon the same expert demonstration data via Dynamic Policy Programming, a stochastic optimal control method. Together these two elements provide a combined assistive mobility system, capable of operating in restrictive environments without the need for additional obstacle avoidance protocols. Experimental results in both simulation and on the University of Technology Sydney semi-autonomous wheelchair in settings not seen in training data show promise in assisting users of power mobility aids.

Emerging Trends in Realistic Robotic Simulations: A Comprehensive Systematic Literature Review

“…• Autonomous Wheelchair [24], [96]- [98] • Rescue and healthcare services in disaster scenarios [99] • Monitoring and visualisation of ROS data in healthcare [100] • Assistance for elderly people in healthcare settings [101] • Human-vehicle interaction in healthcare [68], [102] • Long-term care facilities (LTCFs) [31], [103] • Autonomous pedestrian following (following medical staff) [104] • Medication delivery and vital signs monitoring [105] Warehouse and logistics…”

“…CAMShift 11 [93], HOG 12 + SVM [129], Actor-Critic RL [140], YOLOv3-Tiny + KCF 13 [2], RRT [17], [143], [144], DDPG 19 [15], [145], [146], GFE 20 [147], A3C [148], Dyna-Q algorithm [11], [149], SARL* 22 [150], ROAMFREE [69], Two-stage stochastic programming [81], TEB 23 + A* [95], [127] 2 -Global Path Planning IDE 24 [16], Dijkstra's algorithm [119], [120], [151], SARSA [152], DQL [8], [153], [154],…”

“…LSTM 45 [102] 3 -Collaborative Control ACT-R 46 [97], DTW 47 [29], SMM 48 [124] 4 -Voice Commands Control NLP 49 [28], [175], [176] 1 RBPF: Rao-Blackwellized particle filter 2 EKF: Extended Kalman Filter Navigation serves as an integral component, bridging path planning and control mechanisms by combining varied techniques to ensure robust movement within diverse environments. There are hybrid approaches to navigation, such as the combination of Convolutional Deep Neural Networks (CDNN) and Multi-Goal Reinforcement Learning (MGRL) [165], and the synergy between Artificial Neural Networks (ANN) and Dynamic Policy Programming (DPP) [24]. Algorithms like the Dynamic-Window Approach (DWA) [166] and Deep Reinforcement Learning (DRL) [37] further enhance the robot's adaptability and responsiveness to dynamic scenarios.…”

“…On the other hand, HITL has found a somewhat broader application, although it still presents limited engagement [24], [28], [29], [34], [53], [89], [98], [102], [103], [112], [119], [124], [176]. Most of these studies interpret HITL as a human control mechanism rather than a human-mediated testing method [28], [89], [98], [102], [112], [119], [176].…”

“…Other HITL applications are concerned with human collaboration with robots [29], [124] and social interactions [103]. Only a few pioneering works are leveraging HITL in the robot development process, such as [24], where a robot system learns from user demonstrations and [34] where human feedback trains autonomous methods for unsafe action prediction. Another notable mention is [53], emphasising how human interaction can lead to rapid robot adaptation, eventually eliminating the need for human intervention.…”