The challenge aims to further develop the innovative AI-based solution for Autonomous Guided Vehicles (AGVs) of IntellIoT. The solution should leverage advanced machine learning and computer vision techniques to enable AGVs, such as tractors or forklifts, to navigate, perform tasks, and interact safely within a dynamic environment.
a) One of the main bottlenecks at this point is the lack of labelled training data for control of IntellIoT’s tractor. The main challenge here is to have video feeds associated with the remote-control commands, which require a human operator to manually collect them by driving the tractor around obstacles. Here, the aim is to develop a generative model (video-GAN) that could be used to augment such data streams labeled with accurate control commands to avoid the need for human effort. The augmentation could take into account the adoption of different obstacles as well as different climate conditions.
b) Reinforcement Learning environment: To depart from data collection and labelling process, it is necessary to move towards an RL setting where the learning algorithms require access to agents that interact with the environment. In this view, to develop RL-based control policy for the e-Tractor, an accurate simulation environment that cooperate perception, dynamics, and reward mechanisms is needed. Hence, this task is to develop a simulation environment and learn a control policy using RL.
a) Fleet Management: Create a solution that enables centralized management and coordination of multiple AGVs, optimizing task allocation, resource utilization, and workflow efficiency.
b) Human-Machine Interaction: Design intuitive interfaces or voice commands that allow workers to interact and collaborate with AGVs easily, enhancing productivity and usability.
c) Energy Efficiency: Optimize the AI algorithms and AGV operations to minimize energy consumption and increase the overall sustainability of the autonomous systems.
d) Object Recognition and Manipulation: Further develop the computer vision techniques that allow AGVs to detect, recognize, and interact with objects, such as pallets or crates, and perform tasks like picking, placing, or stacking.
e) Human-Awareness and Safety: Incorporate AI capabilities that ensure AGVs operate safely in the presence of humans, with features like pedestrian detection, collision avoidance, and dynamic environment adaptation.
f) Fault Tolerance and Resilience: Implement mechanisms that enable AGVs to handle unexpected situations, recover from errors, and maintain operation under varying conditions.