Hi,
Will there be again a race and challenge mode next saison?
A time-race would be nice too. Or maybe driving counterclockwise.
regards,
xStatic
Self-driving robot racing with the Raspberry Pi
We are still working out the timetable for the summer series, at the moment nothing is definite yet.
Once we know the plan we will update both the dates and rules pages and make sure everyone knows what is different from the last series. In the meanwhile any thoughts or ideas are appreciated.
In previous Formula Pis we had a timed elimination round just before the final. This was each entry on its own trying to set a best lap time. Is that the kind of thing you are thinking of for a timed race, or did you have something else in mind?
When Claire asked for feedback on the challenge rounds I sat down and wrote a very long analysis about why the challenge rounds could be a good thing and what went wrong with them this season, and lots of suggestions for improving them in the future. But I haven't yet sent it in. This is why:
When Tim and Claire went out drinking the weekend before the final and started tweeting about buying toys for the track i replied with some suggestions for better things to spend money on. The only reply I got back was along the lines of "Waah we're such martyrs. Do it our way or eff off." (Paraphrasing "If you haven't got something positive to say then say nothing."). I'm actually quite offended by that. A company pretending to teach robotics should at least understand the principles of positive and negative feedback.
So actually I'm kind of in a limbo. I have a huge amount of feedback to give you but I am afraid of the reaction that i will recieve for it and I am not sure that I care enough any more to argue about it. I don't see why all of the effort we put into racing (despite the results, sometimes) and all the understanding we have of how the systems work should be disregarded like that. As far as I can tell the whole setup is run for the benefit of Tim Freeburn only and if we were to criticise that then we could easily be expelled and never heard from again. It is obvious that some very basic errors occured this season and I feel a bit cheated, like we've all been used as part of a big experiment without our consent and also without there being any clear benefit to it at all.
So here is the best single piece of advice I can give you...
Send out the competitors satisfaction surveys that we've been promised. Publish the results from the surveys and involve competitors (and other key stakeholders) in the decision making processes. Stop making decisions behind closed doors and be transparent about your motives.
I am having doubts about whether or not to enter the next season. Breakdowns in communication between the organisers and competitors like we've seen this season would be the most likely reason for me to withdraw and probably not return.
Wow! Thanks for the comments Pico.
You're absolutely right, we ARE saying "Do it our way or eff off" as you so politely put it. Our way is that everyone should be nice to the stewards and other competitors, and be positive. There is no need to bring negativity into Formula Pi. If anyone doesn't agree with this, Formula Pi probably isn't for them. However, I do understand this clause isn't in the rule book, so we'll add a rule about positivity for next season. Please take that into consideration before purchasing your next season entry.
I do think a lot of your venom here and in the YouTube chat is due to the changing of the rules after challenge 1. I agree we made the wrong decision up front about not replacing the robots. It wasn't ideal, but it was either change the rules midway or have a season that was unfair to everyone, and would have been painful to watch. You even advocated for change in the YouTube chat. Rerunning the Challenge 1 wouldn't have been fair for those who did well in it. We took the decision to change for challenge 2 and 3.
Based on points, even if you had been awarded full points for the first challenge round, you would not have made it to the final. You would have placed 14th. In the grand scheme of things, it only made a small difference to the overall results which is one of the reasons we decided to do it.
You say we make decisions behind closed doors and aren't transparent about our motives. On the 21st December, we sent out an email to all competitors with the subject "Formula Pi: Challenge 1 logs". Tailed on to the end of that email was a paragraph that we were expecting to change the repositioning rule. There was no response by anyone. On the 22nd Jan. We sent out an email again to all competitors with the subject "Formula Pi: Changes to the challenge rules". That email had three replies - all positive to changing the rules. You didn't respond to it at all. Those so called "closed doors" are very open, but you have to walk through them to get to the other side.
The whole experiment thing? Yes, we are experimenting with Formula Pi and we will continue to do so. We don't know what we are doing, we are learning! The only reasonable way for us to learn is to experiment. There has been only the one rule change this season. Next season we will do things slightly differently based on the input and feedback from this season. If you DO want to Formula Pi to progress, and it's not just about having a go at us, then please do send us your thoughts. I can't guarantee we will do everything you suggest, but we certainly won't ignore them.
In response to the toys that I bought at MCM Comicon, where we sat down and I had two whole 330ml bottles of beer: For some time, we've been looking for and buying decorations for the track with the intention of automating them. Hopefully that will work as a kind of advertisement for PiBorg and thus help us put a little more money into Formula Pi. The cost of improving cameras on the track is orders of magnitude greater than the cost of those toys.
Formula Pi doesn't make us money, it costs us money to run. We do it because we love it. We do this after work in our spare time, so it doesn't get as much time as we would like to give to it. Sometimes because of that we can't be as prepared as we'd like to be.
If you can be positive, we really would like to have you compete next season. The spreadsheet you made for Claire, your simulations and other content was great. You have a great teamname, and have obviously put in effort. It would be really good to see that pay off into results at the end of next season.
Thanks,
Timothy
Yes something like that.
Maybe like in the race heats, 5 competitors having each 2-3 minutes to drive their fastest lap, and then they get points in the order of the best lap times.
Sorry Pico I feel you were a bit harsh on Tim there, they are only trying to do whats best for everyone and I'm not sure they deserved some of those comments. Its meant to be fun, a hobby suitable for all ages and abilities.
I'm pretty sure any re-runs would have a random outcome (not to mention anyone losing points as a result would feel robbed!) - based on some of our races with very similar code but very different lap counts I'd bet you could run a race 5 or 10 times with the same 5 robots and code each time and get a wide range of results and lap counts - it only takes a small change to upset the balance at the moment - I'm sure when the code is more mature there will be more teams who are robust to these changing situations.
I think the rules should be fluid enough mid-season to correct for issues that arise from Tim and Co. trying new things, as long as everyone is aware in advance I don't see the harm - it'd be worse if the whole season didn't work well just because of one bad idea! The challenge rounds were a new addition this year which I personally think are great at promoting innovation and learning for coders of all ages and abilities - as with all these things its going to take a while for the formats and code to develop to best make use of (or cope with) the obstacles.
I've said this before but I think the we need a few seasons with minimal hardware or format changes to allow the code to develop - I find by the time I've developed something its too far into the season (and therefore risky) to try something completely different. I think the two formats, Raptors and Monsters, are a good choice of options and will be good for a few years now.
Regarding the challenge rounds I sent some suggestions to Aaron in an email but I'll post it below and welcome any feedback from the masses:
- maybe keeping the full 5 competitors in a challenge event (perhaps 2 rounds per season rather than 3 though) and
- instead of parked robots have black boxes (so easy to detect) that block a whole lane (or 2 even!), maybe 3 that are heavy enough so they don't move when driven into, and 2 that are light(ish) and will move. This would promote good path planning and track learning for the heavy blocks and also good object detection, avoidance and real time replanning for the objects that are likely to move during the event.
- I'd also like a 10 minute (or more) session to encourage teams to slow down and consider the event rather than just racing round as before. Maybe have some other way of allocating points than just laps completed or fastest run.
Thank you Enigma for your post! I esspacially like the part with the"lets use colored obsticels" That would make the challanges much easier. But I have to add, that the point system should be more dinamic. To get 10 or 0 points(what happen several times last series) only leads to the situation, that some teams are simply out of reach for teams which rather get their points in the race heats. The idea behind the challenge mode is great and as the name allready says"challanging" I like this, but without having several occasions to test different approaches, its really hard to get something running as stable as possible.
Sorry, its another long one!
I completely agree XStatic, its very easy for someone to jump up or down the leader board just based on just 1 or 2 heats, especially as not all competitors can run at once, a scoring system across all cars might work...
Although this idea would have completely ruined us last season what about 1 point per lap completed? Maybe a bonus for winning or coming top 3? Or maybe top 10 to 15 lap counts for the whole week score on a fixed point system. A 5th place result in one heat could get more points than the winner of another just based on lap count and would reward consistency, clean racing and avoidance although it would potentially be at risk of someone else ruining your own race, but I guess this in itself would promote recovery strategies.
Would some longer test sessions with 5+ cars on at once be beneficial, that way teams can test multiple strategies in one session, there are dynamic cars to avoid and there is potential to collect a lot more data. I'd love a free for all 20-30 minute session test session to test 4+ strategies at once and have a good few laps of each to monitor the results.Obviously some cars may not get off the line but these could be restarted a couple of times just to confirm
Finally, the monsters seem quite susceptible to batteries getting dislodged which can be frustrating. the battery packs slung under the robot held in by a single cable tie is risky for a fixed connection point, a short flexible fly lead would be a great addition. Out of interest @PiBorg, can you share any detailed images of the Raptor class cars without the lids on? I assume once they've rolled over they're stuck and cant self-right?
Thanks
Adam
Thanks for the input Adam.
I really like the idea of a point per lap. It's an easy implement, and I do think it makes it a bit easier to swallow when things go wrong. A bonus for the win also sounds like a good idea. We'll discuss the rules and all the suggestions ASAP as we're running out of time to secure the rules and dates. We will most likely be on to both of these points next week.
I don't see a problem with the longer test sessions with multiple cars. Does anyone else agree/disagree with this? A 20 lap test session (no points), but where humans are allowed to interact to right the robots when they get stuck. Great idea!
The battery disconnection isn't an easy fix. We don't really want the general move to LiPo's for safety reasons, and the problem isn't so much the single cable tie. Often the batteries actually disconnect from the sprung connections in the 10xAA holder. The AA packs are foam backed on both sides to hold the batteries in place, but the impacts can be quite brutal. I don't think watching it from cameras gives the right impression, but the robots do an amazing job at holding together especially in a head on crash (which happens quite a lot!). We have discussed getting soldered battery packs (ones with solder tags on them) but it causes a headache for charging, and we'd need to replace the chargers. It will probably remain a pain for a little while longer :(
Post preview isn't working for me, but hopefully I've attached a picture of the Raptor internals.... :S
Thanks Tim for the feedback and info.
I'd be wary of asking for opinion for rules/testing ideas on here though based on previous comments regarding communication and info sharing, I'm not sure how often this page is visited, perhaps once you've collated some thoughts and ideas an email to all competitors might be a good idea to get a some feedback.
Add new comment