0
$\begingroup$

I am a 1st-year grad student majoring in robotics. I have the opportunity to select and execute my own projects for a course called Robot Modeling. For the project, I have decided to model a UR5 and perform trajectory planning. The task would be to pick an object from one place and deploy it to another place while avoiding any obstacles in between. I am constrained to use MATLAB Simscape multibody for my project and it will only be a simulation.

Is the problem statement challenging enough to be completed in 8 weeks?(assuming I can dedicate 5 hours a week). I would love to hear some take on problem statements that would make this more interesting or challenging problem statement for me.

Thanks!

$\endgroup$
  • 1
    $\begingroup$ This sounds like a neat course project - you can set the difficulty by choosing which algorithms to run, how much to implement yourself, and how much uncertainty to include. I would read some relevant literature on RRT* or FMT*, implement the algorithm, and see if you can get it to work when there is noise in the motion model. That said, 40 hours sounds very slim for a course project. $\endgroup$ – combo Sep 21 '17 at 6:47
  • $\begingroup$ Yes, I agree that 40 hours seems slim. But I am constrained to it because I have 2 other courses (Machine learning and control systems) which also have project components. So I have to dedicate 5 hours each per week for 8 weeks for them as well. I will definitely look into the different algorithm to tune the difficulty. Thanks for that! $\endgroup$ – therobotgeek Sep 21 '17 at 17:33
  • 1
    $\begingroup$ I don't know how your school manages it, but I was able to combine project courses from concurrent classes (needed instructor approval). I think you could do a killer project where you model the system well and then provide the model to a basic learning framework to find a good controller. Then you could compare to a more standard control approach to the same task. You should be able to find scenarios that you can learn but which the standard approach fails hard on. $\endgroup$ – combo Sep 21 '17 at 18:18
0
$\begingroup$

Is the problem statement challenging enough to be completed in 8 weeks?

It is hard to tell. In this project, you are tackling several problems at once. All of them requires a solid and at least an advanced-intermediate level of knowledge in physics, math, programming and robotics. For real-case scenario, probably the problems are less in comparison with simulation because you don't need to handle modeling the system. What I would say is 8 weeks are enough but you need extra work than 5 hours (i.e. may be 10). For your case, what I would suggest is to do the following:

1- (Half Week) Finish modeling the system (i.e. requires physics knowledge and half week for this problem is enough). Use Newton's or Lagrangian's approaches. Come up with the differential equations that govern the motion of the system. For saving time, you may consider reading some articles that already done this work with this robot.

2- (Half Week) Finish solving the differential equations (i.e. requires math and in particular numerical method knowledge and programming, matlab will do the hard work for you). Put the ODEs equations in a compact form.

3- (One Week) Choose control approach to move your robot from one point to another. In your case, I will carry out the PD-computed torque controller. It is simple and efficient if you don't have uncertainty in the system, if you do then choose PID-computed torque approach. Perform several tests to make sure that the robot is ready to follow the trajectory efficiently with a reasonable speed. At this stage your robot can move and follow any trajectory BUT it doesn't react with the environment.

4- (Two Weeks) Be acquainted with how the robot interacts with the environment in order to pick up objects (i.e. requires some robotics knowledge). Also, you need to understand how to get data from sensors and let the robot decide for a proper action. For example, when the robot should pick up an object? How does robot know?

5- (Two Weeks) understand obstacles avoidance approaches (i.e. requires some robotics knowledge). Pick up the simplest approach since you don't have an ample time to handle different approaches.

$\endgroup$
  • $\begingroup$ He said that a 'week' only contains 5 hours of work! $\endgroup$ – FooTheBar Sep 21 '17 at 11:52
  • $\begingroup$ @FooBar, I don't care about what he/she said or thought. $\endgroup$ – CroCo Sep 21 '17 at 11:54
  • $\begingroup$ You are answering a question and do not care about what the question is. That's kind of ridiculous. $\endgroup$ – FooTheBar Sep 21 '17 at 11:56
  • $\begingroup$ @FooBar, the ridiculous thing is to build an airbus plane in two days. The completion time the OP suggests depends on several factors. If the OP has no knowledge whatsoever in physics, math, and programming, then this job may take months if not years. If the OP is an expert in this field, it may take 3 weeks give or take. $\endgroup$ – CroCo Sep 21 '17 at 12:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.