Weighted trajectory similarity determination based on sub-trajectory querying
Querying trajectories is disclosed. A database of rich typical trajectories, which includes both position and attribute information for each of the typical trajectories, is created. Using a current trajectory of a node or a current position of the node, the database is queried for a rich typical trajectory that likely corresponds to the current or anticipated trajectory of the node. The database is searched by identifying candidate trajectories and determining a similarity score for each of the candidate trajectories. The best scoring candidate trajectory may be returned and logistics operations may be performed using the best scoring candidate trajectory.
1 . A method for determining an anticipated trajectory of a node currently operating in an environment for logistics operations, the method comprising:
repeatedly:
querying a database of trajectories to select candidate trajectories from the database of trajectories, wherein the database of trajectories includes trajectories of multiple nodes;
extracting comparison points from each of the candidate trajectories to compare with comparison points from a current trajectory of the node, wherein the comparison points from each of the candidate trajectories and the comparison points from the current trajectory relate to positions;
determining a similarity score for each of the candidate trajectories, wherein the candidate trajectories are weighted based on attributes and/or characteristics of the node, and wherein the similarity score includes a discount factor representing a relative importance of past trajectory points;
identifying a final trajectory based on the similarity scores of the candidate trajectories; and
returning the final trajectory to the requestor, wherein the final trajectory is the anticipated trajectory of the node; and
performing the logistics operations in the environment for the node based on the anticipated trajectory of the node, wherein performing the logistics operations comprises controlling a real-world object associated with the node to make decisions in accordance with the anticipated trajectory,
wherein the anticipated trajectory is selected from among a set of rich typical trajectories representing clusters of historical trajectories associated with metadata describing node context, environmental conditions, and task attributes.
2 . The method of claim 1 , further comprising generating the database of trajectories, wherein the database of trajectories include rich typical trajectories, wherein each rich typical trajectory is generated by clustering a set of historical trajectories and associating the resulting cluster with aggregate attribute metadata, and wherein each typical trajectory represents a collection of trajectories that share spatial and/or temporal coherence.
3 . The method of claim 1 , wherein selecting candidate trajectories includes selecting the candidate trajectories based only on positioning data.
4 . The method of claim 3 , wherein the candidate trajectories include typical trajectories that include a point within a threshold distance from a point of the current trajectory of the node.
5 . The method of claim 1 , wherein determining a similarity score for each of the candidate trajectories includes executing a similarity scoring function that considers attribute importance.
6 . The method of claim 5 , further comprising applying a temporal discount factor to weight the similarity scoring function, wherein the discount factor is a decay function applied to historical comparison points based on their age relative to a current position of the node.
7 . The method of claim 6 , wherein the discount factor denotes a time-decay coefficient that reduces the contribution of older comparison points in the trajectory when computing similarity scores, and wherein the decay function is non-linear and based on exponential or logarithmic functions of elapsed time.
8 . The method of claim 1 , further comprising determining anchoring points when a number of the comparison points extracted from a candidate trajectory is different from a number of the comparison points from the current trajectory of the node.
9 . The method of claim 8 , wherein an anchoring point is associated with more than one of the comparison points from the current trajectory of the node.
10 . The method of claim 1 , wherein the logistics operations including decision making operations.
11 . A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations for determining an anticipated trajectory of a node currently operating in an environment for logistics operations, the operations comprising:
repeatedly:
querying a database of trajectories to select candidate trajectories from the database of trajectories, wherein the database of trajectories includes trajectories of multiple nodes;
extracting comparison points from each of the candidate trajectories to compare with comparison points from a current trajectory of the node, wherein the comparison points from each of the candidate trajectories and the comparison points from the current trajectory relate to positions;
determining a similarity score for each of the candidate trajectories, wherein the candidate trajectories are weighted based on attributes and/or characteristics of the node, and wherein the similarity score includes a discount factor representing a relative importance of past trajectory points;
identifying a final trajectory based on the similarity scores of the candidate trajectories; and
returning the final trajectory to the requestor, wherein the final trajectory is the anticipated trajectory of the node; and
performing the logistics operations in the environment for the node based on the anticipated trajectory of the node, wherein performing the logistics operations comprises controlling a real-world object associated with the node to make decisions in accordance with the anticipated trajectory,
wherein the anticipated trajectory is selected from among a set of rich typical trajectories representing clusters of historical trajectories associated with metadata describing node context, environmental conditions, and task attributes.
12 . The non-transitory storage medium of claim 11 , further comprising generating the database of trajectories, wherein the database of trajectories include rich typical trajectories, wherein each rich typical trajectory is formed by grouping multiple historical trajectories using a similarity-based clustering algorithm, and associating aggregate metadata including environmental, operational, and node-specific attributes to the clustered trajectories.
13 . The non-transitory storage medium of claim 11 , wherein selecting candidate trajectories includes selecting the candidate trajectories based only on positioning data.
14 . The non-transitory storage medium of claim 13 , wherein the candidate trajectories include typical trajectories that include a point within a threshold distance from a point of the current trajectory of the node.
15 . The non-transitory storage medium of claim 11 , wherein determining a similarity score for each of the candidate trajectories includes executing a similarity scoring function that considers attribute importance.
16 . The non-transitory storage medium of claim 15 , further comprising applying a temporal discount factor within the similarity scoring function, wherein the discount factor is applied to comparison points based on their temporal distance from a current position in the node's trajectory.
17 . The non-transitory storage medium of claim 16 , wherein the discount factor denotes a time-based weighting function that exponentially reduces the impact of older trajectory points, and wherein the function dynamically adjusts based on real-time operational context.
18 . The non-transitory storage medium of claim 11 , further comprising determining anchoring points when a number of the comparison points extracted from a candidate trajectory is different from a number of the comparison points from the current trajectory of the node.
19 . The non-transitory storage medium of claim 18 , wherein an anchoring point is associated with more than one of the comparison points from the current trajectory of the node.
20 . The non-transitory storage medium of claim 11 , wherein the logistics operations including decision making operations.