Outline Previous Accomplishments o Last year's SURG o Mapkin Proposal Concept o Why is this useful? o The MikroKopter platform o Previous work Criteria For Success
Previous Accomplishments Quadrotor with AeroQuad base Flexible platform for aerial robotics research. Implemented full automatic stabilization o Only inputs required are desired translation and rotation. o PID Tuned from AeroQuad base o Sonar based altitude hold Controller capable of supporting other aerial platforms.
Quadrotor in action Flown by remote control "Stable mode" lets the quadrotor balance itself "Acrobatic mode" is rate control for the quadrotor's movement.
Mapkin Our Build 18 Project, first week of the semester Ground vehicle XBOX Kinect sensor Map out obstacles in a 2D Grid Future goal: SLAM in 2D to learn concepts to apply to Quadrotor
Proposal Concepts "Aerial Point-Cloud Generation using the Microsoft Kinect on an Autonomous Quadrotor" Last SURG: outdoor, unstructured environments, control of the quadrotor and mapping. Current proposal: o Constrained to indoor environments, tight obstacles to avoid o XBOX Kinect depth and color sensing o Point-Cloud generation in 3D (offboard) o More online computation for autonomous navigation Goal: quadrotor navigates itself a short distance indoor, builds a point-cloud representation on a nearby computer
Why is this useful? Feedback from robots needs to be useful for humans Applications: o Exploring disaster zones o Exploring unreachable areas o Military Navigation can be autonomous or RC Point-clouds mean the robot's perspective could easily be understood by a human operator
The MikroKopter Platform Quadrotor kit Many parts of the navigation, etc are done (most are editable code) Waypoint navigation provided We'll need to add: o Microsoft Kinect sensor for vision o Any additional sensors (e.g. downward sonar) o BeagleBoard processor for onboard control
Previous Work CMU Robotics Institute, including the Micro Air Vehicle Lab University of Pennsylvania, "Aggressive Maneuvers" o We'd like our perception and obstacle detection onboard. Our own last SURG o We're confident we can get a flying quadrotor quickly and move on to complex behaviors Other mapping/SLAM projects o
Criteria For Success Successfully fly a Kinect around inside using RC Navigate inside autonomously over a short distance Produce a point-cloud map from Kinect imagery Final Question: Is our product usable?
Questions?