Dongkun Han

Hejun Huang, Zhenglong Li, Dongkun Han

Safety guarantee is essential in many engineering implementations. Reinforcement learning provides a useful way to strengthen safety. However, reinforcement learning algorithms cannot completely guarantee safety over realistic operations. To address this issue, this work adopts control barrier functions over reinforcement learning, and proposes a compensated algorithm to completely maintain safety. Specifically, a sum-of-squares programming has been exploited to search for the optimal controller, and tune the learning hyperparameters simultaneously. Thus, the control actions are pledged to be always within the safe region. The effectiveness of proposed method is demonstrated via an inverted pendulum model. Compared to quadratic programming based reinforcement learning methods, our sum-of-squares programming based reinforcement learning has shown its superiority.

Dongkun Han, Dimitra Panagou

Estimating the Domain of Attraction (DA) of non-polynomial systems is a challenging problem. Taylor expansion is widely adopted for transforming a nonlinear analytic function into a polynomial function, but the performance of Taylor expansion is not always satisfactory. This paper provides solvable ways for estimating the DA via Chebyshev approximation. Firstly, for Chebyshev approximation without the remainder, higher order derivatives of Lyapunov functions are used for estimating the DA, and the largest estimate is obtained by solving a generalized eigenvalue problem. Moreover, for Chebyshev approximation with the remainder, an uncertain polynomial system is reformulated, and a condition is proposed for ensuring the convergence to the largest estimate with a selected Lyapunov function. Numerical examples demonstrate that both accuracy and efficiency are improved compared to Taylor approximation.

Dongkun Han, Lixing Huang, Dimitra Panagou

We consider the multi-task coordination problem for multi-agent systems under the following objectives: 1. collision avoidance; 2. connectivity maintenance; 3. convergence to desired destinations. The paper focuses on the safety guaranteed region of multi-task coordination (SG-RMTC), i.e., the set of initial states from which all trajectories converge to the desired configuration, while at the same time achieve the multi-task coordination and avoid unsafe sets. In contrast to estimating the domain of attraction via Lyapunov functions, the main underlying idea is to employ the sublevel sets of Lyapunov-like barrier functions to approximate the SG-RMTC. Rather than using fixed Lyapunov-like barrier functions, a systematic way is proposed to search an optimal Lyapunov-like barrier function such that the under-estimate of SG-RMTC is maximized. Numerical examples illustrate the effectiveness of the proposed method.

Kunal Garg, Dongkun Han, Dimitra Panagou

This paper presents a robust distributed coordination protocol that achieves generation of collision-free trajectories for multiple unicycle agents in the presence of stochastic uncertainties. We build upon our earlier work on semi-cooperative coordination and we redesign the coordination controllers so that the agents counteract a class of state (wind) disturbances and measurement noise. Safety and convergence is proved analytically, while simulation results demonstrate the efficacy of the proposed solution.

Ning Liu, Sen Shen, Zheng Li et al.

Hierarchical multi-robot exploration commonly decouples frontier allocation from local navigation, which can make the system brittle in dense and dynamic environments. Because the allocator lacks direct awareness of execution difficulty, robots may cluster at bottlenecks, trigger oscillatory replanning, and generate redundant coverage. We propose VORL-EXPLORE, a hybrid learning and planning framework that addresses this limitation through execution fidelity, a shared estimate of local navigability that couples task allocation with motion execution. This fidelity signal is incorporated into a fidelity-coupled Voronoi objective with inter-robot repulsion to reduce contention before it emerges. It also drives a risk-aware adaptive arbitration mechanism between global A* guidance and a reactive reinforcement learning policy, balancing long-range efficiency with safe interaction in confined spaces. The framework further supports online self-supervised recalibration of the fidelity model using pseudo-labels derived from recent progress and safety outcomes, enabling adaptation to non-stationary obstacles without manual risk tuning. We evaluate this capability separately in a dedicated severe-traffic ablation. Extensive experiments in randomized grids and a Gazebo factory scenario show high success rates, shorter path length, lower overlap, and robust collision avoidance. The source code will be made publicly available upon acceptance.

Dongkun Han, Dimitra Panagou

Formation control is a key problem in the coordination of multiple agents. It arises new challenges to traditional formation control strategy when the communication among agents is affected by uncertainties. This paper considers the robust multi-task formation control problem of multiple non-point agents whose communications are disturbed by uncertain parameters. The control objectives include 1. achieving the desired configuration; 2. avoiding collisions; 3. preserving the connectedness of uncertain topology. To achieve these objectives, firstly, a condition of Linear Matrix Inequalities (LMIs) is proposed for checking the connectedness of an uncertain communication topology. Then, by preserving the initial topological connectedness, a gradient-based distributed controller is designed via Lyapunov-like barrier functions. Two numerical examples illustrate the effectiveness of the proposed method.