Institution profile

Jimei University

Academic institutionasia · cn
Official website
Research library8linked papers
Opportunities0open roles
Selected work

Representative Papers

Multi-UAV Formation Cooperative Obstacle Avoidance and Adaptive Shape Deformation Control in Complex Environments Based on BI-APF-RRT and Affine Transformation

Jun 28, 2026

This work addresses the challenge of simultaneously ensuring obstacle-avoidance agility and formation integrity for multi-UAV systems operating in complex environments. To this end, a cooperative planning framework is proposed that integrates a target-biased bidirectional artificial potential field–RRT (BI-APF-RRT) with affine transformations. The BI-APF-RRT algorithm generates smooth, rapidly converging global collision-free trajectories while circumventing the local minima commonly encountered in conventional artificial potential field methods. Concurrently, affine transformations incorporating non-uniform scaling and rotation enable adaptive, dynamic reshaping of the formation along the planned paths. A distributed control law further ensures coordinated navigation through cluttered spaces. Experimental results demonstrate that the proposed approach effectively balances collision avoidance safety with formation coherence, significantly enhancing the autonomous navigation capabilities of multi-UAV systems in complex scenarios.

0 citationsRead paper

Representation over Routing: Overcoming Surrogate Hacking in Multi-Timescale PPO

Apr 15, 2026

In multi-timescale reinforcement learning, naively aggregating rewards under different discount factors often leads to surrogate objective misuse and myopic degradation. This work proposes a Target Decoupling architecture that preserves multi-scale value predictions in the critic to enhance representation learning while restricting the actor’s policy updates exclusively to long-horizon advantages, thereby preventing interference from short-term signals. By decoupling the policy and value update pathways, the approach effectively circumvents these pitfalls. Implemented within a PPO framework and augmented with multiple discount factors, auxiliary representation learning, and gradient isolation mechanisms, the method consistently surpasses the “solved” threshold on LunarLander-v2, significantly improving performance, eliminating policy collapse, and escaping local optima inherent to single-timescale approaches.

0 citationsRead paper
Recent publications

Latest Papers

Multi-UAV Formation Cooperative Obstacle Avoidance and Adaptive Shape Deformation Control in Complex Environments Based on BI-APF-RRT and Affine Transformation

Jun 28, 2026

This work addresses the challenge of simultaneously ensuring obstacle-avoidance agility and formation integrity for multi-UAV systems operating in complex environments. To this end, a cooperative planning framework is proposed that integrates a target-biased bidirectional artificial potential field–RRT (BI-APF-RRT) with affine transformations. The BI-APF-RRT algorithm generates smooth, rapidly converging global collision-free trajectories while circumventing the local minima commonly encountered in conventional artificial potential field methods. Concurrently, affine transformations incorporating non-uniform scaling and rotation enable adaptive, dynamic reshaping of the formation along the planned paths. A distributed control law further ensures coordinated navigation through cluttered spaces. Experimental results demonstrate that the proposed approach effectively balances collision avoidance safety with formation coherence, significantly enhancing the autonomous navigation capabilities of multi-UAV systems in complex scenarios.

0 citationsRead paper

Representation over Routing: Overcoming Surrogate Hacking in Multi-Timescale PPO

Apr 15, 2026

In multi-timescale reinforcement learning, naively aggregating rewards under different discount factors often leads to surrogate objective misuse and myopic degradation. This work proposes a Target Decoupling architecture that preserves multi-scale value predictions in the critic to enhance representation learning while restricting the actor’s policy updates exclusively to long-horizon advantages, thereby preventing interference from short-term signals. By decoupling the policy and value update pathways, the approach effectively circumvents these pitfalls. Implemented within a PPO framework and augmented with multiple discount factors, auxiliary representation learning, and gradient isolation mechanisms, the method consistently surpasses the “solved” threshold on LunarLander-v2, significantly improving performance, eliminating policy collapse, and escaping local optima inherent to single-timescale approaches.

0 citationsRead paper