I-Corps: Advanced simulation system for end-to-end autonomy validation in robot vehicle systems

The broader impact/commercial potential of this I-Corps project is in simulation technology that can be used for validation and optimization of any autonomous robot vehicle system. This technology’s most significant benefit is for robots (Uncrewed Air, Ground, Marine Vehicles) in sectors involving challenging operation environments including defense, space, transportation, delivery, construction, energy, mining, and agriculture.

Unlocking rapid and reliable deployment at scale is key for the manufacturers and service providers active in these sectors and the limitations of their current method for test and validation can become a significant obstacle for deployment at scale. Current processes for validation include field tests, lab experiments or using disintegrated software tools to test different parts of the system in isolation.

Lack of large-scale end-to-end test data that would enable them to rapidly validate the autonomy stack of these robots can result in an autonomy that cannot be relied upon eventually leading to increased operation inefficiencies, manual labor, risks and costs in development and resources.

This I-Corps project is based on the development of an advanced modular simulation system to validate, enhance, and optimize the autonomy stack in autonomous robots end-to-end. It provides scalable testbeds for robot autonomy through a connected modular data-driven simulation where the robots’ software, hardware, and artificial intelligence can be tested through all phases of operation in multi-fidelity virtual spaces and challenged in different realistic scenarios.

Using a novel approach in multi-modal data collection, through a performant rendering mechanism and a hybrid modular design for underlying engines, the simulator generates large-scale test data to validate and optimize the components in autonomy stack for single or multi-robot scenarios before deployment thus mitigating risk of deployment and accelerating validation. The underlying modular engines provide physically accurate sensor and object modeling, data-driven large-scale performant environment modeling, software and hardware in-the-loop testing, human feedback integration and predictive vision modeling for integrated test and training of machine learning models and other software and hardware components in the autonomy stack.

By accelerating validation by a factor of 100 to 100,000, this technology can enable built-in resiliency through reliable autonomy stacks that are optimized and resilient against interference or sources of errors.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.