How can you make sure when your flying robot still comes back to its desired point and perform stable flight, even when it is hit in the face (or even kicked)?
Here is a video from IDA-PBC controlled aerial robot preserving its stable and desired flight in a hursh human-robot physical interaction. Video shows the stablilizng power of passivity-based controllers for aerial physical interaction task. Thanks to IDA-PBC one can further tune the physical behavior of the system in such scenarios.
Controlling aerial robots in a stable way is already a challenge. Imagine the complexity of controlling the aerial robots equipped with manipulating arms. In this work we present a differential flatness based control method, which takes the nonlinear dynamic coupling of the complex system into account, and can be implemented using off-the-shelf setups e.g. common servo motors, in real-time.
Imagine flying robots that can physically interact with their environment, for purposes e.g. surface inspection or cleaning.
It is certainly not an easy task, since achieving a stable flight for a robot is still an open topic in real world scenarios.
In this video, we try to solve this problem by designing a light-weight flexible joint arm, which can achieve intrinsically safe aerial physical interaction, thanks to its passive compliance.
It is not far from today, that these robots can fly to a construction yard and help building a structure, e.g. by doing hammering. We humans use the compliance of our muscles to amplify the velocity of our arms to maximize the efficiency of hammering task. Why not these flying robots cannot do the same, using light-weight flexible joint arms?
Human-robot interaction is not a new topic, and previous studies show that the collaboration of both can certainly increase the performance of task. Here we control the flying hammer with our own hand, thanks to the Leap Motion hand tracking sensor.
We develop our own codes and frame works, from the exact velocity control of the propellers, to the high level control of the robot. Hence everything in our system is transparent to us. Here we hack the pitch control of the quadrotor, to track a hand motion of a human. We track the hand motion using Leap Motion sensor. The delay is due to the algorithm that smoothens the Leap Motion data up to fourth derivatives (four time differentiable).