Project
| # | Title | Team Members | TA | Documents | Sponsor |
|---|---|---|---|---|---|
| 26 | Computation, localization, and power systems for distributed robotics |
Lyle Regenwetter Tanitpong Lawphongpanich |
Jonathan Hoff | design_document3.pdf final_paper1.pdf final_paper2.pdf other1.pptx |
|
| Members: Lyle Regenwetter (regnwttr): EE-ME double major, responsible for the power/propulsion and localization systems of the hovercraft. Tanitpong Lawphongpanich (tl6): CE major, responsible for the computing systems and localization systems. Problem: We have been working with Professor Dullerud on distributed networked robotics for a while. Professor Dullerud’s research focuses on coordinated multimodal drone swarms, and he tasked us with the development of an autonomous hovercraft for use in his research. We found that currently no commercially available robotics systems are suitable for use on the hovercraft. The older generation of these hovercrafts are over 12 years old and each use a hacked-together drone controller and a raspberry pi along with a Vicon localization system, which all in all makes for an unreliable and inflexible setup. The localization system is also costly, needs to be run in a precalibrated environment, and needs line of sight to the robot, which is not ideal for multiple agent robot swarms. More info about the Hovercraft project (2004 paper, current one don’t have published paper yet) :https://drive.google.com/open?id=1LG_8ff0SxGruOKPeisfWHJae3BoWNj18 Note that the current versions of the drones are newer than the one in the paper which uses drone controller and Vicon instead of custom boards and cameras like in the paper. The theory part about state estimation and air bearing remains the same. Overview: We plan to build a computing, localization, and power systems for distributed robotics which is low cost and able to adapt to various types of mobile robots. We will use an ultra wide band and gyro/magnetometer for the localization, while the computing systems will have a linux computer and a microcontroller to handle real-time tasks, the power systems should allow battery hot-swapping. Solution components: 1)Computing system: This will consist of a master Linux single board computer (raspberry pi compute module) and a salve micro-controller board (ideally a 32-bit ARM microcontroller). The linux computer will be responsible for high-level tasks such as trajectory generation and some math related computation on numpy or related python packages. The micro-controller will be used for time-critical tasks such as controlling a PID loop, outputting PWM to control actuators, reading sensor data and some low-level digital filters. Both will communicate through serial or i2c bus. This will include a wifi-adapter to connect to a network as well. 2)Localization systems: The robot needs to know its location, i.e. x-y coordinates as well as its orientation in the world frame. We are planning to use magnetometer+gyroscope for absolute orientation, and gyroscope to increase the accuracy through a fusion algorithm. For the x-y-z position, we are planning to use one or multiple decawave ultra-wide-band chips which gives out the time of flight of the wave from each anchor on the wall, combined with a Kalman filter to increase the accuracy. We are hoping to have an absolute accuracy of within +- 15 cm (roughly the size of a soccer ball) 3)Power systems: The power system provides 5V dc and 3.3V DC for the computing/localization systems with protection from back EMF from the motors and actuators. The battery should be hot-swappable, which means that we can swap the robot's battery while it is running without the need to shutdown the computing systems that we might have been running an experiment. Project goals: We expect to have every sub components laid out on PCBs by the end of the semester. We plan to prioritize the computation and localization system, which means that we can reduce the scope of the power systems if we don't have time. Currently, we have some computation and localization systems prototyped on breadboards with jumper wires. It should not be too hard to just migrate and polish these two subcomponents on printed circuit boards and design the new power system in one semester based on the old one. Currently, the mechanical and propulsion part of the hovercraft is almost done and we have the hovercraft running with the prototype breadboard hardware. What we will be doing and what are we going to demo: We will be designing and testing the PCBs for the systems during the semester. We will demo at the end of the semester to show that we can get the orientation and location data in the world frame from the sensors (with our implementation of fusion algorithm and Kalman filter running on the compute unit) and be able to control motors using the board we build, all of which will be running on the custom power board that we designed this semester too. Successful demo shows that we can obtain the precise location and orientation of the robot and be able to drive motors using the systems we design and have some python code reading the sensors value/output PWM through the micro-controller. |
|||||