Elijah H. W. Ang presents Evolving Drone Bodies and Minimal Perception for Autonomous Navigation
On 2025-09-25 - 2025-09-25 14:00:00 at G205, Karlovo náměstí 13, Praha 2
Note: To accommodate more participants, the seminar has been rescheduled from
11:00 to 14:00 in room G205. It will still take place on the same day.
The MRS group will host Elijah H. W. Ang, a postdoctoral researcher
under Guido de Croon at the MAVLab, TU Delft. He will present his current
research on Evolving Drone Bodies and Minimal Perception for Autonomous
Navigation. You are cordially invited to attend the seminar and discussion.
Original Elijah’s description:
The design of multicopter drones has remained largely unchanged since their
inception. While conventional designs perform adequately across a wide range of
applications, they may not be the optimal design for specific tasks. Drawing
inspiration from nature, we look to artificial evolution as a method to explore
and optimize drone designs for navigation tasks.
In the context of autonomous navigation with obstacle avoidance, high-fidelity
sensors such as stereo cameras or LiDAR are often employed for their
robustness.
However, such sensors are often bulky, heavy, and demand substantial onboard
computation, making them impractical for small and agile drones. In cases when
artificial evolution produces highly compact and nimble morphologies, such
sensing strategies are no longer viable.
In this talk, I will present our current research which aims to revisit drone
design by exploring which body morphologies and sensor choices are optimal for
different tasks through artificial evolution. This talk comprises of two main
topics: (1) joint evolution and reinforcement learning of drones for various
navigation tasks, and (2) evaluating and implementing minimal perception
strategies with reinforcement learning for navigation and obstacle avoidance.
11:00 to 14:00 in room G205. It will still take place on the same day.
The MRS group will host Elijah H. W. Ang, a postdoctoral researcher
under Guido de Croon at the MAVLab, TU Delft. He will present his current
research on Evolving Drone Bodies and Minimal Perception for Autonomous
Navigation. You are cordially invited to attend the seminar and discussion.
Original Elijah’s description:
The design of multicopter drones has remained largely unchanged since their
inception. While conventional designs perform adequately across a wide range of
applications, they may not be the optimal design for specific tasks. Drawing
inspiration from nature, we look to artificial evolution as a method to explore
and optimize drone designs for navigation tasks.
In the context of autonomous navigation with obstacle avoidance, high-fidelity
sensors such as stereo cameras or LiDAR are often employed for their
robustness.
However, such sensors are often bulky, heavy, and demand substantial onboard
computation, making them impractical for small and agile drones. In cases when
artificial evolution produces highly compact and nimble morphologies, such
sensing strategies are no longer viable.
In this talk, I will present our current research which aims to revisit drone
design by exploring which body morphologies and sensor choices are optimal for
different tasks through artificial evolution. This talk comprises of two main
topics: (1) joint evolution and reinforcement learning of drones for various
navigation tasks, and (2) evaluating and implementing minimal perception
strategies with reinforcement learning for navigation and obstacle avoidance.