A Method for Hierarchical Control of Mobile Robots with Pairwise Kinematic Constraints

Velocity Obstacles (VO) and Optimal Reciprocal Collision Avoidance (ORCA) are widely used for local collision avoidance in multi-agent robotics, but they typically assume point- or disk-shaped agents without kinematic coupling. This paper extends ORCA to pairs of holonomic robots that are rigidly connected by a shared payload, modeled as moving segments. Each agent’s desired velocity combines goal progress with inter-agent distance maintenance; the latter is enforced via a radial velocity penalty in a Model Predictive Control (MPC) layer. To handle collisions with moving segments formed by other coupled pairs, we introduce virtual non-cooperative ORCA agents placed at the closest point on the segment, with velocity linearly interpolated from the endpoints and a cooperation factor of η = 1. The hybrid VO–ORCA–MPC framework preserves decentralization and scales linearly. Simulations with 10 robots demonstrate zero collisions and a formation RMSE of 13.5%. The average speed of coupled pairs exceeds that of single agents by 20.64% due to the non-cooperative treatment of virtual agents.