Decentralized Multi-Robot Warehouse Surveillance with Orientation-Aware Planning and Hybrid Collision Avoidance

The surveillance of critical or hazardous indoor environments — warehouses, industrial facilities, and convenience stores — is often too risky or expensive for human operators, motivating coordinated multi-robot teams. This paper presents a decentralized pipeline for grid-based multi-robot surveillance: a density-weighted utility function chooses high-priority targets, an orientation-penalized variant of A* plans turn-efficient routes to them, a Linear Quadratic Regulator(LQR) tracks the resulting trajectories, and a three-tier collision-avoidance scheme — escalating from artificial potential fields to grid-aligned shape-navigation boundary tracing to a bounded local A* replan — coupled with a time/cost-based negotiation protocol keeps robots clear of static obstacles, dynamic obstacles, and each other. We demonstrate that the orientation-aware cost function preserves A* optimality using 2 different robot sample sizes. The orientation-aware planner decreased in-place turning by 70.19% over 40 randomized trials, matching or beating baseline A* in all 40 of the 40, and by 64.85% (N = 15) / 59.11% (N = 25) on the real targets utilized throughout, cutting real execution time under LQR monitoring by 25.13% (N = 15) / 21.60% (N = 25). Collision-avoidance escalation resolved 15 of 15 (N = 15) and 25 of 25 (N = 25) dynamic-obstacle encounters and negotiation resolved all 6 genuine timing conflicts found across both team sizes (1 of 105 pairs at N = 15, 5 of 300 at N = 25). The results show that the pipeline is a feasible and scalable method to multi-robot surveillance of congested interior settings, with the target-selection mechanism validated for team sizes from 5 to 30 robots and the full pipeline verified at two representative scales, N = 15 and N = 25.