“UAM will make large cities feel small, with quick rides across town in autonomous, electric vertical-takeoff-and-landing aircraft,” said Wei. “The potential of UAM comes with big operational challenges, though, that can only be solved using sophisticated methods from control, optimization, artificial intelligence and machine learning.”
Wei’s team is creating algorithms and models for UAM operations from takeoff to touchdown:
- Arrival management models schedule landings based on both promised arrival times and how much battery power each UAM aircraft has left – and will help design the most energy-efficient arrival trajectories.
- Collision avoidance seeks to mimic the human intelligence of people walking in a crowded room without bumping into each other. Algorithms will provide flight-path-prediction capability to avoid crashes.
- Autonomous air traffic control combines centralized intelligence on the ground with distributed intelligence in the air for maximum UAM capacity and safety.
- Airspace management balances demand with available airspace and flightpaths to answer key questions about how many UAM aircraft can take off at a time, from where and for how long, given the variable conditions of weather and available electric battery on each aircraft.
- Fleet dispatch tools will optimize UAM dispatch, so companies can do precise and efficient scheduling of their fleet. Models will also integrate passenger and cargo uses for UAM.
Wei’s UAM operations models are built using industry-provided data sets and tested in both open-source computer simulations and controlled experimental simulations in an unmanned aerial vehicle lab on campus.
“UAM models are incredibly complex. Proving to people that something as never-done-before as UAM is safe and reliable will take a lot of careful, incremental validation and working closely with government officials to write operations policy,” said Wei.