24 Jun

RoCo: extend the behavior to enable more analysis of our designs

Chang Liu computational design

RoCo demonstration brainstorming

We came up with some challenging problems for RoCo demo and we will keep brainstorming. (Thank you Nabeel and Mauricio) Here are some ideas (currently focusing on Paperbot, will extend to other robots later on):

  1. Robot going through large ridges or bumps: robot will get stuck on a ridge if wheel diameter is not large enough or if the robot is too wide.

  2. Robot following path with sharp turns: smaller wheel diameter is needed to follow precisely, larger wheel diameter is needed to finish the task on time.

  3. Robot going up or down a very steep incline: need certain wheel diameter to maintain speed and torque, certain mass to control friction, but large wheels will cause the robot to tip.

  4. Battle bots/ramming another robot: need high mass but low friction to acceleration to a high momentum (optimizing momentum, wheel radius is also a factor) [I need more time to think about this one, it sounds interesting, but I need to think through]

Project overview

I also talked to Pehuen, we agreed on how this research will be carried out and Pehuen will help with the optimization part of the problem. The basic logic is (this is also for Nabeel to catch up with the progress): We have RoCo take design parameters (e.g. w, l, h, etc) as input, expand the functionality of RoCo to generate 3D ready-to-simulate design file (the more I think about it, the more I feel like we need generate a file can be applied not just to Webots but other simulation tool as well, so I think .obj will be our best option.) Once we have the automatic generated file, we will apply that to Webots + some solid environment (no Scenic involved at this stage). Output from Webots (e.g. v) will be grouped with RoCo input (e.g. w, l, h, etc) as a set of input-output-relationship. We will be some (maybe 10-100) sets needed in order to build an input-output-relationship so that we can give the best redesign suggestion from performance. I will work on automating RoCo to generate ‘simulatable’ robots, and Pehuen will work on the optimization part later on.

I have also talked to Yusuke regarding my current progress, so that Yusuke can carry on the Webots simulation part of the problem.

RoCo implementation questions

  1. In Paperbot.yaml, the paperbot is considered as one piece. Is this understanding correct?

Paperbot.yaml

  1. In function STLWrite(faces, fp, kwargs), the faces contain all parts, body and wheels. Do you know if there is an easy way to differentiate the body and wheels? (I know this is do-able since it is separated in DXF 2d drawing, but I am having trouble finding where it happens)

  2. Can we split different body part in the ESPSegBuilder.py? We will still have a whole body ESPSeg.yaml saved to library in the end, but adding different components along the way, e.g. body.yaml, wheel.yaml. Do you think this is a valid approach?

Next Post Previous Post

Search

Popular Tags

foray origami robots blimp ducky vr edg SLAM computer vision computational design roco slam Origami rocolib arduino RoCo design Robot Compiler simulation Motion Planning autonomy edg-model create; kalman filter; mesh reinforcement learning; multi-agent; mean-field; logics imu contacts mesh network motors esp32 friction networking tank dynamics pushing auxetic esp now machine learning multi-agent visual processing communication wireless control i2c webapp ROCO pcbs structure path planning computer skill cars openmv Robotica reinforcement learning power consumption modeling origami dataset test visual odometry surface structure dash port-arrays Battery DARPA motor gondola machine vision controller semantic paper auxetics infrastructure feedback gestural interface frequency diversity mixed reality robot Consensus bistable auxetic unit cell gestural intuition crawler darpa; create; logic create; kalman filter; EM; structures Foray network open house propeller lidar sensors apriltags 433mhz optimization Quantum Information 4-wheeled FORAY perception control environmental design refactoring icra2021 radio Kullback-Leibler Divergence crazyflie DER Auxetics lasercutter distributed optimization blynk gate von Mises Fisher Distribution pushing problem experiment results or nodemcu offline reinforcement learning exam research proposal conference object detection simulink webots project management modeling contacts tutorial thrust IR motor configuration integration test localization UAV legs joystick altimeter decentralized ros2 balloon test painlessmesh mesh networking swarm vision processing arnhold matlab fusionbot RL arduino nano multi-agent path planning gazebo PCB rviz2 nonholonomic systems Integrated Demo reinforcement learning; slam; paper review vision Short Post trajectory optimization edg-examples serial monitor python Model-Free Machine Learning Q-Learning rssi controls feedback-controlled grant writing system ID multirobot system signal strength quasi-static pushing marginal cost pricing in-situ design IROS neural network hot plates cell reinforcement learning; multi-agent; mean-field; game theory; mean field; reinforcement learning current reinforcement learning; multi-agent; Xbee Zigbee robot compiler origami robot deployment spi dynamic modeling rhino boat literature review multipack buoyancy equations contact surp paper review Writing motorized boats throughput range surfaces ML reservoir computing multi-agent path finding solar cars detectionrate distance shape color edg-ide pain Characterization tool concept surface contact liquid crystal elastomer conferences release peripheral device differential keyboard swarm control car von Mises-Fisher von Mises Fisher port-array quasi-static modeling Kalman Filter Zero Elastic Energy modes Liquid Crystal Elastomers Discrete Elastic Rods Consensus Algorithms power pivoting Learning 3D-CHESS; kinematic_model elastic polymers 2.4 GHz 433 MHz creative Computer skill internal misc edg-refactoring 433 ceres debug coriolis force structured field dynamic model proposal wtf deep learning interpolation wireless signal strength reinforcement learning; multi-agent; decentralized reinforcement learning; multi-agent; mean-field; stationary demo video flytrap papers steering plastic multi-object tracking pvdf prelab actuation von Mises Distribution Barycenter Multi-Agent Systems integration flapping color_segmentation center of friction lab_data result segmentation on multiple object and centers finding competition ZMP buoyancy current sensor multiple blimps battery pid defender Synthesizable Radio integrated system 3d printing catching mechanism kalman filter barometer detection balloon slam; flowchart; depth estimation grid chiral square unit cell state space multi-robot pressure distribution model prediction thrust configuration kalman filter; ESP32 paper cutter COVID 19 manipulation fabrication code hardware offline Rl firmware protobuf #activecapturingmechanism #claw ensemble methods Bayesian networks 3D-CHESS; kalman filter bootstrapping model-based RL boxboat tracking lift force cad nylon scenic adaptive control motor driver MIMO communications Euler-Bernoulli beam newton's method spring-mass submission kalman filtering point cloud 2.4ghz latency der codesign sma-based actuator presentations machine training cameras preliminary exam multiple devices inkscape material delta tolling semantic segmentation instance segmentation pdms silicone optitrack multirobot solidworks ICLR design tool board design HDL pixracer pixhawk ardupilot flight controller regression fly test motor distribution Python DNN ICRA game theory ICML guassian processes probabilistic neural networks graph generators MBRL ide batch learning ros sensor fusion von Mises-Fisher Distributions edg-compiler Bessel functions iot