Obstacle-Aware IBVS Target Tracking Via Feature-Space Projection and Virtual Imaging Guidance with ADP-Shaped Terminal Cost
Mingcong Li, Zhen Chen, Xiangdong Liu
AI summary
Problem
Vision-guided wheeled robots struggle to simultaneously track targets and avoid obstacles in mapless environments due to underactuation, kinematic coupling, and gimbal-only local minima that stall navigation.
Approach
The method decomposes control tasks via weighted feature-space projection to allocate tracking to the gimbal and navigation to the base, generates safe heading references through virtual imaging constraints, and learns context-aware terminal costs online via approximate dynamic programming.
Key results
- Feature-space projection resolves underactuation and prevents gimbal-only local minima
- VICG generates visibility-preserving heading references for mapless obstacle avoidance
- Online ADP learns context-aware terminal costs for adaptive long-horizon MPC guidance
- Hardware experiments validate real-time feasibility and effective joint tracking/avoidance
Why it matters
Enables safe, vision-only navigation for wheeled mobile robots in unstructured environments without pre-mapped data, advancing autonomous inspection and search-and-rescue applications.
Abstract
We propose a visual-servoing and obstacle- avoidance controller for a wheeled mobile robot (WMR) with a two-axis gimbal camera that operates without mapping, using only vision and lightweight forward sensing. A task- allocation MPC with online terminal-cost iteration is in- troduced. Specifically, task projection in the image-feature space mitigates underactuation and coupling–induced local optima; Virtual Imaging Constraint Guidance (VICG) yields a visibility-preserving heading reference that steers the trajectory around obstacles; and an Approximate Dynamic Programming (ADP) module learns a context-aware terminal cost online, providing long-horizon guidance for mid-horizon prediction. Relying solely on image feedback plus lightweight ranging, the method coordinates the WMR and gimbal to accomplish obstacle avoidance and visual-servo tracking jointly. Hardware experiments validate the feasibility and effectiveness of the proposed approach.