AI Perception System Crash

The Perception System crashes when exiting PIE on UAIPerceptionSystem::UnregisterSource.
It doesn’t check for null sense classes which are there by design I suppose.
I transition maps and some of them have AICharacters with different senses so senses not registered are ZeroInitialized in UAIPerceptionSystem::RegisterSenseClass causing their corresponding Senses[SenseID] == nullptr.


  • Could you please provide your logs from your project’s Saved->Logs folder?
  • Could you cause the crash again and provide your Machine ID from the Crash Reporter window?
  • Do you have any steps we can take to reproduce this issue on our end?


I am marking this topic as resolved for tracking purposes, as we have not heard from you in a few days. If this issue persists, feel free to respond to this thread. For any new issues, please create a new Answerhub topic.

Have a great day

this issue is not resolved and if you don’t even bother to report it as a bug or whatever then I see this as a failure of Epics support. In my opinion if a person responsible for this code or any skilled programmer would see my description he would immediately identify the problem. I am doing my best to give an extensive description of my problems so a programmer familiar with how the system works won’t need any more information. Until today Mieszko Zielinski who always answered to my AI requests/bug reports never needed any more info and was very helpful in a timely manner.

Hi Marek,

It’s standard procedure for us to mark an AnswerHub post as resolved when we do not get a response to the questions we ask after 4 days. You failed to respond in any way to Sean’s questions, so we marked it answered. This allows us to properly track issues that have active responses and maintain a manageable work load.

It does not mean that you can’t provide the requested information or that we won’t see it if you do so. If you would like us to continue investigating, please feel free to answer Sean’s questions and we’ll move forward.

Our engine developers have a limited amount of time, and cannot provide personal, individual support. Our support team is here to verify and pass along bug reports to them. If we are not given the information we need to properly reproduce and report an issue, it will go nowhere. Please take another look at the questions Sean asked, and we’ll go from there. Thanks.

We’re not offended. If you’d like to provide the information we requested, that would be helpful. Without it, we’ll go ahead and resolve this post until we have more to work with. Thanks for your time.

If you’ve already fixed this issue (or any others), you may prefer to submit a Pull Request. It’s Epic’s preferred workflow for integrating fixes from programmers such as yourself. Each of these are run by the developers to see if they want to integrate the fix or create an alternate version of the fix that suits them better, and it’s a good way to discuss it with the developers as they get to the request.

For now, we can run this by Mieszko, but any of the additional information we requested may help us expedite this process.

Hi Ben,

Sorry if I offended you, this wasn’t my intention. A crash is a major issue for the engine so I feel it should be in your interest to fix it. My company is working on a game so we don’t have time to do QA for you. I fix engine stuff myself since we can’t wait for new releases anyway so when I send you a crash report I do it in good faith, hoping that you can improve the engine for your customers. I’m just disappointed seeing that you ignore it. As a programmer i’d rather get a clear description of an error than a pure callstack and a bunch of logs. What I’m trying to say is that the procedure should be more flexible and should account for such situations. As I stated in my previous message I’ve had good experience with my prior questions where the answer was more constructive and I NEVER attached any callstack/logs and nonetheless the error was recognized and fixed almost immediately on master branch.