Bidirectional search
Finds the quickest route from one vertex to another by searching forward from the start and backward from the goal at once, until the two searches meet.
Time complexity O((V+E) log V), gograph function path.Bidirectional.
How it works
Bidirectional search runs two Dijkstra searches at once: one forward from the start and one backward from the goal. Each step takes the vertex with the smaller time from the two queues, so the two sides grow at the same pace. Each side covers about half the distance, so on a road map it looks at a fraction of what one search would. Routing engines use it on its own and under faster methods such as contraction hierarchies.
The search from the goal follows edges in reverse: on a directed graph it walks the edges that lead into a vertex, so a one-way street that leads into the goal is taken backward from the goal. Each time one side reaches a vertex the other side has reached, the two times add up to a route, and the search keeps the quickest one it has found.
It can't stop the moment the two sides first touch, since a quicker route can pass through vertices neither side has settled. It stops once the two smallest times left in the queues add up to no less than the best route: any route not found yet goes through a vertex still in a queue and takes at least that long.
path.Bidirectional takes the edge cost as a function, and a nil cost uses the edge weight. A negative cost returns ErrNegativeWeight, and a goal the start can't reach returns ErrUnreachable. Ties break by the order vertices and edges were added, so the same graph gives the same route on every run.
Downtown grid
Twenty intersections on a downtown street grid. Streets go both ways, Market is closed to cars between 2nd and 3rd, and each edge is the travel time in minutes.
A directed, weighted graph with 20 vertices and 60 edges.
What is the quickest drive from Pine & 1st to River & 5th, with Market closed between 2nd and 3rd, searching from both ends at once?
The quickest drive from Pine & 1st to River & 5th takes 15 minutes, along 7 streets.
Step by step
- Search forward from Pine & 1st and backward from River & 5th at once. Each side has its own queue, ordered by time from its end, and each round settles the vertex with the smaller time of the two.
- The next times are 0 from the start and 0 from the goal: 0 + 0 = 0, and no route is known yet, so go on. Settle Pine & 1st from the start at 0.
- Pine & 1st to Pine & 2nd: 0 + 2 = 2.
- Pine & 1st to Main & 1st: 0 + 2 = 2.
- The next times are 2 from the start and 0 from the goal: 2 + 0 = 2, and no route is known yet, so go on. Settle River & 5th from the goal at 0.
- Market & 5th to River & 5th leads into River & 5th, so the search from the goal takes it backward, from River & 5th to Market & 5th. It follows every street in reverse, so a one-way street counts only the way it leads to the goal. 0 + 2 = 2.
44 more steps lead to the last one:
- Stop: the next times, 9 from the start and 7 from the goal, add up to 16, no less than the best route, 15, so no route left to find can be quicker. The quickest route meets at Pine & 4th and takes 15 minutes. The searches settled 10 vertices from the start and 7 from the goal.
Use it in Go
route, minutes, err := path.Bidirectional(g, "Pine & 1st", "River & 5th", nil)
if err != nil {
return err // path.ErrUnreachable when the goal is out of reach
}
fmt.Println(minutes)
for _, v := range route {
fmt.Println(v.Label())
}
Example graphs
- New York streets, 295 vertices and 669 edges
- London streets, 274 vertices and 598 edges
- San Francisco streets, 248 vertices and 548 edges
- Rasht streets, 298 vertices and 568 edges
- School run, 6 vertices and 13 edges
- Delivery area, 8 vertices and 12 edges
- Downtown grid, 20 vertices and 60 edges
- One-way city, 45 vertices and 88 edges
- Metro region, 100 vertices and 346 edges
More in Shortest paths
- Dijkstra: Finds the shortest distance from a start vertex to every other vertex when no edge weight is negative.
- A*: Finds the quickest route from one vertex to another, looking first where a guess of the time left says the goal is, so it settles far fewer vertices than Dijkstra.
- k shortest paths: Finds the k quickest routes from one vertex to another that don't repeat a vertex, quickest first, the way a navigation app offers a few routes.
- Beam search: Looks for a route from one vertex to another while keeping only the few best partial paths at each step, so its memory and time stay bounded, at the cost of sometimes missing the best route.
- Multi-source Dijkstra: Finds the nearest of several sources for every vertex, and the distance from it, in one search.
- k-center: Chooses k centers so that the vertex farthest from its nearest center is as near as it can make it.
- Bellman-Ford: Finds the shortest distance from a start vertex to every other vertex, even with negative weights, and reports a negative cycle when the start can reach one.
- Floyd-Warshall: Finds the shortest distance between every pair of vertices, even with negative weights, by letting paths stop at one more vertex in each round.