search

RoboJackets / robocup-software

Georgia Tech RoboJackets Software for the RoboCup Small Size League
Apache License 2.0
183 stars 187 forks source link

Replan just the last segment when there is a goal change #264

Closed ashaw596 closed 6 years ago

ashaw596 commented 9 years ago

Should be able to implement a very simple direct path for small goal changes. Add the plan onto the end of the path.

justbuchanan commented 9 years ago

@ashaw596, when we talked about this yesterday, we discussed implementing this by adding some sort of function to Path for extracting a subpath. One point you brought up was whether it should return a new Path object or modify the original. I think it might be good to make it return a new one b/c it would make it easier to move path-planning to its own thread at some point. We would leave the original untouched, then swap to the new path as soon as it's done being processed on the path-planning thread.

justbuchanan commented 9 years ago

Any update on this one?

justbuchanan commented 9 years ago

I guess we'll probably need the new path-planner in place for this to happen... That part may not happen for comp this time around since there are several other things that take priority that need work (firmware mostly).

ashaw596 commented 9 years ago

Yeah it needs the new planner in place unless I fix the other planner. What do we need to do with firmware? Or can I finish the new planner? On Jun 30, 2015 3:08 PM, "Justin Buchanan" notifications@github.com wrote:

I guess we'll probably need the new path-planner in place for this to happen... That part may not happen for comp this time around since there are several other things that take priority that need work (firmware mostly).

— Reply to this email directly or view it on GitHub https://github.com/RoboJackets/robocup-software/issues/264#issuecomment-117359034 .

justbuchanan commented 9 years ago

It'd be probably be best if I finish up the new planner since I've kind of got it in a weird state right now :/ Do you have an MBED on you by chance?

ashaw596 commented 9 years ago

=/ I didn't take one with me to cali. On Jun 30, 2015 5:17 PM, "Justin Buchanan" notifications@github.com wrote:

It'd be probably be best if I finish up the new planner since I've kind of got it in a weird state right now :/ Do you have an MBED on you by chance?

— Reply to this email directly or view it on GitHub https://github.com/RoboJackets/robocup-software/issues/264#issuecomment-117377136 .

ashaw596 commented 9 years ago

Lol I guess I could see if anyone in my company happens to have one. Probs not tho. On Jun 30, 2015 5:18 PM, "Albert" ashaw596@gmail.com wrote:

=/ I didn't take one with me to cali. On Jun 30, 2015 5:17 PM, "Justin Buchanan" notifications@github.com wrote:

It'd be probably be best if I finish up the new planner since I've kind of got it in a weird state right now :/ Do you have an MBED on you by chance?

— Reply to this email directly or view it on GitHub https://github.com/RoboJackets/robocup-software/issues/264#issuecomment-117377136 .

jgkamat commented 9 years ago

I'm pretty sure I brought one here, I could give it to you if you needed it!

justbuchanan commented 9 years ago

Ok, there's probably not a whole lot you can do with the firmware then :/ How much work do you think it would be to setup the replanning to replan only part of the path if I am able to get the new planner in place in time? If it's not too much, I'd say go ahead and do it, and hopefully I can get the planner done.

justbuchanan commented 9 years ago

Yeah if you guys could arrange that, that would be awesome! Not sure how far you guys are from each other. I expect that the majority of our time at comp will be spent frantically trying to get firmware up and running, so it definitely wouldn't hurt to be familiar with it.

ashaw596 commented 9 years ago

Hey jay! Yea that be awesome. Can I grab that mbed sometime? Where do you live? On Jun 30, 2015 5:42 PM, "Justin Buchanan" notifications@github.com wrote:

Yeah if you guys could arrange that, that would be awesome! Not sure how far you guys are from each other. I expect that the majority of our time at comp will be spent frantically trying to get firmware up and running, so it definitely wouldn't hurt to be familiar with it.

— Reply to this email directly or view it on GitHub https://github.com/RoboJackets/robocup-software/issues/264#issuecomment-117379424 .

jgkamat commented 9 years ago

I'll send you an email!

justbuchanan commented 9 years ago

@ashaw596, you were asking earlier about the parameters to the main planner method. The planner isn't 100% finalized, but I don't think it'll change from this:

std::unique_ptr<Path> plan(const Geometry2d::Point& startPos, const Geometry2d::Point& startVel, const Geometry2d::Point& endPos, const Geometry2d::Point& endVel, const MotionConstraints& motionConstraints);
ashaw596 commented 9 years ago

@justbuchanan ok. Sounds good. Also can we refactoring motion constraints so that it makes more sense with our current planning system. The targetpos and targetworldvel don't make that much sense now that we're also setting a target velocity in for motion planning. I think it would be best to either make a seperate motion commands structure. Or create functions like setspeed or set pathplanning to make its use clearer and add a seperate path planning target vel and change the current one to like "overrideTargetVel" or something

ashaw596 commented 9 years ago

It's just weird the mix of both motion constraints and motion commands is confusing.

ashaw596 commented 9 years ago

Also, in the interface, ur probs missing the obstacles.

ashaw596 commented 9 years ago

Are we doing angle planning in the path btw?

justbuchanan commented 9 years ago

Yeah I agree - let's split up MotionConstraints into one that's actual constraints (velocity and accel limits, etc) and one that's the command. And good call, the set of obstacles will be in there as well. The paths will ignore angles, I think it's probably best to handle that separately.

ashaw596 commented 9 years ago

btw, Can you help me with some boost python stuff? I'm trying to refactor the RObot method to use a unique_ptr to storethe path, but the boost python won't let it compile because it tries to export the RObot's copy method. Theres something like boost:noncopyable you can put in the class template, but I don't really get boost python @justbuchanan

justbuchanan commented 9 years ago

Just downloaded it - I'll take a look and see what I can figure out. I'm not sure why it needs the copy constructor though, we should only be passing pointers to Robots into python land, not making any copies.

justbuchanan commented 9 years ago

Got it fixed! You were right, adding boost::noncopyable fixed the problem. By default, boost::python tries to make a copy constructor for the wrapped class, which in this case wasn't going to work. Apparently we ran out of template arguments for the wrapper, so I removed the one for shared ptrs to OurRobot since we weren't using it and put noncopyable in its place. It's on your albert/replan branch now.