Abstract:
The continued high number of fatalities associated with trackless mobile machines in South Africa’s mining industry has led to the introduction of collision avoidance system regulations in the Mine Health and Safety Act in 2015. These regulations have engendered the profusion of technologically immature collision avoidance systems from third-party vendors; all of which are centred on automatic stopping and braking systems. These braking systems often result in trivial or ineffective solutions, proving costly to mining operations. This study presents a novel collision prediction and trajectory generation model that incorporates the addition of steering control to current collision avoidance systems. The proposed collision prediction model employs a probabilistic methodology that enables the development of opportune trajectories and control objectives. This model’s integration of non-linear state estimation, point-wise contact points, and time-to-collision approximations provides the trajectory generation model with detailed insights to synthesize safe, predictable, and efficient trajectories. The trajectory generation model proposed incorporates a novel Monte Carlo Lattice Hyper Sampling path planner and velocity profiler which is designed to overcome the foresight and convergence shortcomings in many modern path planners. The collision prediction and trajectory generation models were simulated and evaluated using various Earth Moving Equipment Safety Round Table (EMESRT) interaction scenarios. The collision prediction model ensured zero potential collisions went undetected, yet, at times, still triggered the intractable problem of false alarms. The proposed novel trajectory generation model avoided all potential collisions encountered and improved operational efficiency when compared to current braking-only solutions. The addition of steering control and a coupled collision prediction model significantly improved the safety and efficiency of collision avoidance systems in surface mining environments.
