In　view　of　the　current　difficulty　in　setting　PID　controller　parameters　and　high　cost　and　power　consumption,　an　adaptive　PID　system　based　on　embedded　devices　is　designed　by　taking　ball　on　plate　system　as　an　example.　The　article　first　introduces　the　PID　algorithm　and　improves　the　PID　algorithm.The　model　is　then　analyzed　and　simplified　for　further　analysis.　Based　on　the　simplified　model,　a　neural　network　based　on　real-time　parameter　updating　is　proposed　to　improve　the　control　accuracy　and　reduce　the　control　time.　After　simplifying　the　control　network,　it　is　transplanted　to　the　embedded　device　to　implement　PID　adaptive　control.　Finally,　the　feasibility　of　the　system　is　verified　by　comparative　experiments.　©　2019　IEEE.