The near mid-air collision of three jets this week at Washington National Airport illustrates the complexity of collision avoidance systems. A disastrous collision between three jets carrying 192 passengers was narrowly averted by an air traffic controller with only 1650 yards of separation remaining between planes screaming toward each other at a combined speed of 436 MPH. This near disaster highlights something that is easily overlooked about our unconscious mind: how our brain constantly monitors our surroundings automatically and acts to prevent collision as we weave effortlessly through throngs of pedestrians and circumvent obstacles in our path without giving it a thought.
We are all familiar with the experience of approaching another person and instead of passing each other gracefully, you both move abruptly in the same direction putting you on a collision course. Instantly you veer in the other direction but the approaching person makes the same dodge and you come to within a split second of colliding head-on. You both halt instantly, releasing a little laugh.
We laugh for two reasons: such navigational dilemmas and near misses are rare, and secondly, all of this behavior and maneuvering is carried out unconsciously. It's amusing because both of you are just spectators.
But how? Collision avoidance is an extremely difficult problem. Enormously complex systems, sophisticated communication mechanisms, and strict adherence to standard protocols guard against collision in any transportation system. Yet in walking around freely there are no rules of the road for collision avoidance. If we followed a stereotype rule to safely negotiate passing another pedestrian on foot—say, always pass on the right, for example, there would be no problem, but there are no such rules. You and the person approaching you are free to pass each other either on the right or the left. There is no requirement to communicate or negotiate your preference. If left to chance this would seem to end in near collisions 50% of the time, but in fact collisions between people are infrequent. How can this be explained? Does each person somehow read the other's mind and know which way to pass? Without the slightest conscious thought our hidden cerebral guidance system solves this problem instantly and constantly so that you are free to engage your conscious mind in other matters. Researchers Anne-Helene Oliver and colleagues at the University of Rennes in France wondered how this was possible, and they devised an experiment to uncover the secret.
The experiment they devised models the situation where two people walking on a sidewalk but hidden from each other around the corner of a building suddenly find that their paths will intersect at 90 degrees as they emerge from behind the corner, threatening a T-bone collision. When this happens at least one person must divert their path or alter their pace to allow the other person to cross either in front or behind the other. This situation of collision at a blind intersection was reproduced in the laboratory by using a wall to obscure the two subjects as they unwittingly approach. This allowed the investigators to determine precisely how much time the subjects would have to avoid a collision after seeing one another emerge from behind the wall and to videotape the evasive actions each person made.
Their first finding, published in the current issue of Gait and Posture, confirmed what previous research had shown: that people use a different strategy for collision avoidance depending on whether the obstacle ahead is an inanimate object or a human being. If the obstacle to be avoided is an inanimate object, regardless of whether it is stationary or moving, we simply divert our course (and speed) onto a new trajectory to safely avoid colliding with it. If the object ahead is a person, though, this strategy is not used, simply because the other person can suddenly change and move in any direction—speed up, slow down, stop, step right or left. The "steering a safe trajectory" approach is the strategy we use while driving a car to avoid a doggy squirrel in the road, and we all know how badly that can end.
After analyzing video footage of subjects avoiding collisions in their experiments, graphing all the data and reducing it to mathematical equations, the researchers found one mathematical variable that would accurately predict the collision avoidance maneuvers we perform to prevent "pedestrian road kill." The factor was not a variable unique to either pedestrian's behavior, such as a change in speed or direction, it was a factor that combined the actions of them both—a reciprocal interaction of both people—mean predicted distance (MPD). This is the anticipated distance of separation between the two people at the point ahead where their current trajectories would bring them to intersect with each other. If that predicted distance of separation is less than one meter, both subjects altered course in a way to increase the distance of predicted separation at the intersection to one meter or more. If the MPD was greater than 1 meter, neither person altered their stride or direction.
The researchers found that this collision avoidance between pedestrians was achieved in three phases: an observational phase, a reaction phase, and a regulation phase. The mutual guidance maneuvers during these phases are executed in seconds or fractions of a second. The observational phase begins when the two walkers first see each other, and realizing that they are on a collision course (MPD less than 1 m), begin evasive action. This observational phase is only 1/3 of a second, remarkably quick considering all of the situational awareness and subconscious calculation, analysis, and decision making that must be triggered. Next the walkers begin maneuvers to coordinately reduce the MPD if necessary, a period of only 3 seconds. In the final regulation phase, the two walkers fine tune their trajectories and speed to allow them to pass each other even closer than the minimum distance of 1 m clearance the instant they pass. This adjustment occurs in the final 0.8 sec before the crossing point. This is about the time it takes to make one stride. Both walkers can now permit the clearance distance to tighten because the collision avoidance problem has been solved and the other person's trajectory can no longer change in the final instant before passing.
Never, in any of the staged potentially disastrous interactions did the experiments result in a collision. We are pretty darn good at this. We subconsciously calculate what the future distance of separation at a distant point of intersection will be between us and the other pedestrian, and if that value fails to meet our minimum safe distance of 1 meter—about the distance of one stride—both walkers mutually enter into a series of course corrections until both of our on-board computers calculate that the result will be 1 m of clearance; we stick to that course, and we pass.
Oliver, Marin, Cretual, and Pettre (2012) Minimal predicted distance: A common metric for collision avoidance during pair wise interactions between walkers. Gait and Posture 336: 399-404.