달력

09

« 2010/09 »

  •  
  •  
  •  
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  •  
  •  
Abstract :

The ability to simultaneously localize a robot and accurately map its surroundings is

considered by many to be a key prerequisite of truly autonomous robots. However, few

approaches to this problem scale up to handle the very large number of landmarks

present in real environments. Kalman filter-based algorithms, for example, require time

quadratic in the number of landmarks to incorporate each sensor observation. This

paper presents FastSLAM, an algorithm that recursively estimates the full posterior

distribution over robot pose and landmark locations, yet scales logarithmically with the

number of landmarks in the map. This algorithm is based os an exact factorization of

the posterior into a product of conditional landmark distributions and a distribution over

robot paths. The algorithm has been run successfully on as many as 50000 landmarks,

environments far beyond the reach of previous approaches. Experimental results

demonstrate the advantages and limitation of the FastSLAM algorithm of both simulates

and realworld data.




이 논문은 FastSLAM 알고리즘을 제안한 논문이다. 현재 FastSLAM 2.0 까지 나와

있는 상태이며 SLAM문제를 이론적으로 이상적으로 해결한 알고리즘중 하나 이다.

위의 논문을 요약해논 문장에서 보면 진정으로 자유로운 로봇이란 로봇의 위치 측정

이 되는 동시에 주변을 고려한 정확한 맵을 생성하는 로봇이라 정의 하였다. 이말은

SLAM문제를 해결한 로봇을 의미한다. 칼만 필터 기반의 알고리즘은 실제 환경의 수

많은 지표에 대해서 2차 방적싱의 시간이 필요하며 이로 인해 지표가 많은 실제 환경

에서는 사용하기가 곤란하였다. 이에 기하 급수적으로 늘어나는 계산 시간을 줄이고

로봇의 위치와 지표의 위치를 완전한 경험적 확률 분포로 재귀적 평가가 가능한

FastSLAM 알고리즘을 제안한다. FastSLAM의 계산법은 조건부 지표분포와 로봇의

경로의 곱을 경험적 확률로 나누어 정확히 인수 분해 하는 방법을 사용한다. 이로 인

해 이전의 접근법을 넘어서는 50000개 정도의 많은 지표에서도 성공적으로 동작하였

으며 모의 실험과 실 세계에서의 극한의 한도 실험에서 모두 FastSLAM알고리즘이

유익함을 증명한 논문이다.




로봇 알고리즘과 이에 관해서 접해온지 어언 1년이 넘었다. 날로 발전해가는 기술과

알고리즘은 따라가기에도 벅찰 정도로 눈부시게 발전하고 있는 상황이다. 컴퓨터 학

도로써 여러가지를 배워야 하고 전자공학적 지식도 필요한 분야가 로봇 분야 인것 같

다. 나는 아직아직 멀었나 보다. ㅡ,.ㅡ;;;

크리에이티브 커먼즈 라이선스
Creative Commons License
Posted by Tyamo
Abstract :

This paper presents an analysis of the extended Kalman filter formulation of
simultanouse localization and mapping. We show that the algorithm

produces very optimistic estimates once the 'true' uncertainty in vehicle

heading exceeds a limit. This failure is subtle and cannot, in general, be

detected without ground-truth, although a very inconsistent filter may

exhibit observable symptoms, such as disproportionately large jumps in the

vehicle pose update. Conventional solutions - adding stabilising noise,

using an iterated EKF or unscented filter, etc- do not improve the situation.

However, if "small" heading uncertainty is maintained, EKF-SLAM exhibits

consistent behavioour over an extended time-period. Although the

uncertainty estimate slowly becomes optimistic, inconsistency can be

mitigated indefinitely by applying tactics such as batch updates or

stabilising noise. The manageable degradation of small heading variance

SLAM indicates the efficacy submap methods for large-scale maps.



이 논문은 EKF-SLAM의 단점에 대해서 실험을 거쳐서 불확실성과 비효율성을 증명

한 논문입니다. 요약을 조금 줄여보면 EKF-SLAM은 매우 불확실한 한계를 넘어선

알고리즘이며 불균형하며 매우 큰 시간 주기로 갱신이 된다. 전통적인 해결법 보다는

안정적인 알고리즘 즉 개선은 되었지만 단점이 많은 알고리즘입니다. 이의 개선법으

로는 느린 측정과 불일치는 일괄 갱신이나 안정적인 잡음으로 완화가 가능하며 작은

주제의 변화 SLAM의 다루기 쉬운 성능 저하는 큰맵을 위한 작은 맵을 그리는데 효

과적이라는 이야기가 쓰여져 있습니다. 영어 해석이 잘못된것이 있을수 있으나 큰

의미는 틀리지 않은것 같습니다. 결론적으로는 관계 부분맵 결합 방법을 사용하여

속도적 개선이 가능하다는 의미인것 같습니다. 이 알고리즘을 바탕으로 졸작을 해야

하는 입장에서 FastSLAM보다 알고리즘적 효율성이 떨어지나 개선 방법이 있다니

저에게는 반가운 이야기인것 같습니다
크리에이티브 커먼즈 라이선스
Creative Commons License
TAG EKF-SLAM, SLAM
Posted by Tyamo