Matthew J. Roorda is a Professor of Civil Engineering and has been faculty at the University of Toronto since 2005. Dr. Roorda is a Canada Research Chair in Freight Transportation and Logistics, and Founding Chair of the Smart Freight Centre, established 2019. He is Chair of the TRB committee AT015 Freight Transportation Planning and Logistics.
Matthew Roorda and his students Farah Ghizzawi, Ruowei Li, Tho Van Le, and Usman Ahmed worked with Purolator to develop a simulation tool for evaluating the performance of a person-following delivery robot in dynamic pedestrian environments. The prototype is the EffiBOT, a courier-following bot deployed by Purolator. The EffiBOT assists foot couriers in performing last-mile delivery tasks in public areas, such as underground pedestrian walkways and malls. Although similar robots have proved successful in uncrowded spaces like warehouses, it is crucial to evaluate their performance in crowded, more complex environments. Ultimately, this study aims to provide quantitative evidence for lawmakers and logistics companies to make informed decisions on the safe deployment of this technology in public spaces.
First, laboratory experiments are performed to observe the robot’s obstacle avoidance and person-following behaviours. The experiments are designed to accurately reflect a wide range of intersections between the robot and the foot courier it must follow, as well as other dynamic and static obstacles it must avoid. Second, the trajectory data resulting from the experimentation is used to calibrate a social force model, an analytical force-based model that depicts the dynamics of pedestrian movements in response to each other and the environment. This model largely dictates how the robot would move in a simulation model. Third and lastly, the robot’s performance is evaluated in a simulated indoor pedestrian environment where different crowding scenarios are tested. The simulated environment comprises of a pedestrian walkway, and the agents include the robot, its operator, and other pedestrians.
In conclusion, this study presents an effective tool to reproduce the behaviours of person-following robots and evaluate its performance in the context of different environments and under varying crowding scenarios. Although reducing a complex robotic system to a simple social force model may overlook certain aspects of the robot’s behaviour, it still provides a rigorous approach to assist decision makers. Future research endeavors for this study include applications in fully autonomous delivery robots, pedestrian hotspots and safety, and standards and regulations development for the safe deployment of delivery robots in indoor and outdoor urban areas.