A MethodtoEstimate Destination ofaWalking Person withHidden MarkovModelforSafety of HumanFriendly Robots
Soichiro Morishita, Akihiro Nishimura, Hajime Asama · 2008
Inthis paper, weproposed amethodtoestimate behavioral traits aregiven attention(6). Theseareapproaches thedestination ofwalking persons fromtheir walking patterns, for basedonthedefined person's walking model, whichisbased avoidance ofcollision accidents between pedestrians androbots in oncertain assumptions about humanbehavior. Inshort, these aHuman-Robot Coexistence Environment. We adopted theHidden Markov Model(HMM)asamodeltorepresent walking patterns. We models donotexactly reflect anactual pedestrian's behavior. It constructed amodelforeachmovement pattern. A movement pattern isexpected that theaccuracy willbeimproved withmodeling wasdefined withadeparture point anddestination point ofaperson.using anactually observed movementtrajectory according to Comparing thelikelihood withtheachieved model, wediscriminated theenvironment. walking patterns forwhichthedestination isunknown. Wedidsome Basedon theaboveunderstanding, we takethetrajectory experiments inanactual environment toverify theavailability of theproposed method. Results showthatthediscrimination ratioofanobject person usinga fixed camera, andconstruct a approached 80%within 2sofobservation. learning modelofthewalking pattern described using the HiddenMarkovModel(HMM).Then, we define movement