Development of a simulation environment for UGV type robot, which makes it possible to rapidly test control algorithms, robot design, performance, and train AI system using realistic scenarios.

In this research and development effort, simulations are required mainly for two reasons:

1. to test and troubleshoot any developed software algorithm;
2. to train deep learning networks.

Due to the inherent differences in simulation software and its capabilities, two separate approaches will be used to cover all technical aspects.

To simulate the UGV in life-like off-road scenarios with accurate physical terrain, VBS3 will be used whereas for full system design. The specific tasks involve putting together developer and end-user requirements, developing reconfigurable UGV model, creating models for relevant sensors with programmable characteristics, creating interfacing modules, scripting scenarios, and supporting behavioural optimization.

Expected outcomes will be configurable model of the UGV with simulated sensors, interfacing module for training networks and ADM software and specific use-case simulation scenarios in the simulation environment.

During the project simulated environment will be setup. Sensor data for a minimal set of sensors will be simulated. Environment to generate training data for object and terrain recognition will be developed as well as the data infrastructure will be specified. Simulated sensor data for all sensors and simulation scenarios based on use-cases for demonstration will be tested and virtual environments will be developed for end-user testing and training.