When a Couple Goes Together: Walk along Steering

More realistic, flexible, and expressive social crowds using transactional analysis

“…In the latter, the agent that is further ahead waits for the return of the one behind. As noted by Popelová et al [27], stopping and waiting generally looks more realistic than heading backward to reunite. Additionally, we have found that it looks unnatural for an agent that gets ahead to start and stop suddenly as the lagging agent's distance falls in and then out of the formation nucleus.…”

“…This results in agents who are really good friends with only a specific small set of agents and who have no relationship to the remaining agents. For example, agents in Popelová et al's work [27] and Karamouzas and Overmars's work [18] only have relationships with a specific small set of other agents. In this case, SocialReward could be implemented as follows:…”

“…Popelová et al [27] present research focused purely on two people meeting and then walking together to a destination. The agents involved wait for each other when needed and walk abreast by using a social forces-style approach.…”

“…2, focus on particular social situations. For example, Musse and Thalmann [20] model agents in a museum-like environment, Popelová et al [27] focus on situations where people are paired off on dates, and Karamouzas and Overmars [18] look at small groups with an emphasis on shoppers at a mall. Each one of these algorithms is successful in its goals, but none of them is able to handle a large range of different social environments.…”

“…Previous work has moved agents so they can interact by manually altering a specific obstacle avoidance algorithm (for example, social forces [27] or a velocity approach [18]). Since each obstacle avoidance algorithm has its respective advantages and disadvantages, users of these previous crowd algorithms are not at liberty to choose the obstacle avoidance algorithm that Figure 3 shows the abstraction of previous approaches that wrap their implementation around a specific algorithm (top) as compared to our approach, which leaves the decision about the obstacle avoidance algorithm to the user (bottom).…”