No abstract

No abstract

A key feature of an autonomous vehicle is the ability to re-plan its motion from a starting configuration (position and orientation) to a goal configuration while avoiding obstacles. Moreover, it should react robustly to uncertainties throughout its maneuvers. We present a predictive approach for autonomous navigation that incorporates the shortest path, obstacle avoidance, and uncertainties in sensors and actuators. A car-like robot is considered as the autonomous vehicle with nonholonomic and minimum turning radius constraints. The results (arcs and line segments) from a shortest-path planner are used as a reference to find action sequence candidates. The vehicle's states and their corresponding probability distributions are predicted to determine a future reward value for each action sequence candidate. Finally, an optimal action policy is calculated by maximizing an objective function. Through simulations, the proposed method demonstrates the capability of avoiding obstacles as well as of approaching a goal. The regenerated path will incorporate uncertainty information.

This paper presents the architecture of autonomous material handling vehicles. The centralized coordination of multiple vehicles and three-layer architecture (deliberative, sequencing, and reflexive layers) are adopted. The navigation controls, including configuration control, visual servoing, path tracking, and collision avoidance, are developed. The finite state machine (FSM) that supervises the control modules to complete the material handling task is elucidated. The experimental results of a forklift transporting a pallet from an initial to a desired goal configuration are demonstrated.

No abstract