16 May

Cut-off strategy means to use only first N observations to do polyak averaging, and stop updating landmark after N observation.The cut-off strategy has two advantages:More robust than traditional polyak averaging in both global observation model and intermittent observation model Able to...

09 May

This week I've been revisiting the online EM for Hidden Markov Model [1]. The formulation between HMM and latent model is similar but there's a little different. I've also developed an performance evaluator to quantify and optimize the performance of our framework, especially the choice of learnin...

29 Apr

Over the past week, I've successfully transforming the range-bearing measurements to direct observation with correct statistics (without approximation) Now the algorithm is fully functional.Here is the demo setting:maximum range radius: 25 cm range uncertainty: 0.5 cm^2 bearing uncertainty...

18 Apr

Improved algorithm with direct observation: Debugging on why range-bearing measurements failed: Attempted algorithm with range-bearing (perfect rnb observation, noisy time update): Attempted algorithm with range-bearing (noisy rnb observation, perfect time update): Still, problem aris...

11 Apr

For previous two weeks I was building up the framework of EM-SLAM. The framework is showed as following:And as well as the demo by using direct observation (states can be directly observed with gaussian noise):This demo works properly as expected (although there are some discrepency betwee...

12 Mar

Attached is a simple demo on one-step update of EM-SLAM algorithm. Note that the robot has only one observation at the begining, and then with an EM algorithm, the landmarks will converge to a certain configuration, but not exact position due to the prior guess of robot state has error. With a corre...

06 Mar

This week I was trying to derive the mathematical formulation of one-step update for recursive EM-SLAM algorith. I was referring to both batch EM-SLAM paper as well as recursive EM paper for reference, and I figured out a coarse representation of Q-function. While I am still doubt on how to get rid...

20 Feb

This week I was mainly surveying on puplications for recursive EM methods (not restricted to SLAM) in various applications, which is listed below: (1) Online EM algorithm for latent model This paper is more like a tutorial on how to use recursive EM algorithm to estimate parameter and update latent...

30 Jan

To debug what causes the divergence of EM-SLAM algorithm, I Separated E-step and M-step. To prove M-step is correct, every time I will feed a true Q-function from E-step to M-step. i.e. x=x_true, P=0. (Note in general, Q-function contains trajectory information and it is always estimated in E-step)....