Today everyone uses the internet everywhere with smartphones emitting signals for Wifi, Bluetooth, and so on. The history of communication between Bluetooth Low Energy (BLE) becons and smartphones would be beneficial to provide customized application services in order to ehnance user experience in real daily life. For example, the location of a smartphone can be estimated by Received Signal Strength Indicators (RSSIs) of BLE beacons, which could lead to developping a variety of personalized and localized services.
What's next ? The challenge is to predict "where you will be in future" and "where you were in the past", instead of "where you are now", by using the history of locations (or trajectory) of the smartphone.
The task of this competition therfore is to specify start and goal locations, to which the person have departed and is reaching.
BLE beacon trajectories were collected at Nagoya Institute of Technology (Nitech), Aichi, Japan, in Nobember 2017.
Over 150 people participated a game event (called Spy Game) for finding hidden locations by using a smartphone app (called Spyca) tailored for this event. The app uses hundreds of BLE becon signals placed in the campus to estimte the location of the smartphone, but shows the user only the distance to the hidden goal. Hence users wandered the campus for finding hidden goals while seeing the current distance to the goal.
In the figure below, there are 9 locations (green), and the start (blue) and goal (red) of the trajectory are shown. The trajectory is shown with small dots, with the blue circle at the starting point and the red circle at the end point. This kind of trajectories are given as training samples. The task is to predict which of 9 locations are start and goal for test trajectories (that are randomly trimmed in the beginning and the end).
Grant-in-Aid for Scientific Research on Innovative Areas from 2016 to 2021
Navigation is a fundamental behavior of animals including human. In navigation, the following three functions are required: the acquisition of dynamically-changing information from external and internal environment, the choice of route and destination based on the information, and the behavioral regulation to reach the destination. We aim for systems science of bio-navigation to understand the “algorithms” for the navigation of animals. To this end, we bring together experts from control engineering, data science, animal ecology, and neuroscience, and jointly work on how to measure, analyze, understand, and verify bio-navigation.
By hosting competitions, we hope to boost our understanding the algorithms for animal and human navigation.
Please visit our website.
You may submit 5 submissions every day and 50 in total in the development phase.
Submitted results are evaluated by weighted accuracy. As test trajectories are trimmed in its beginning and end with random time period (5 to 60 seconds), prediction becomes harder for longer trimming. Therefore, the period of the trajectory is the weight.
More specifically, wi is weight of i-th trajectory, and pred_si and pred_gi are predicted start and goal locIDs, and gt_si and gt_gi are corresponding ground truth locIDs, of i-th trajectory.
where int(true)=1 and int(false)=0. It is 0 if all predictions are wrong, while 1 if perfect.
There are 234 test trajectory, and classification results should be a 234-line text file, and each line contains two digits (from 0 to 8) for each of start and goal locations of the test trajectory of the line.
Each line has two-digit prediction of the corresponding test trajectory; that is, line 0 is the prediction of start and goal of test trajectory 000.csv. Each of two predicted location IDs (start and goal), separated by a comma, is one digit corresponding to goal_info.csv. If the trajectory 000.csv is predicted as it starts from locID 5 and arrives to locID 3, then the line is "5, 3".
Here is an example:
5, 3 0, 2 1, 3 0, 1 2, 1 8, 0 7, 6
For each trajectory, predict start and goal location IDs from pre-defined locations
A single CSV file (000.csv, 001.csv, ...) contains a trajectory, and each line represents the information of a location of a smartphone. In addition to longitude and latitude, some other information is provided;
The first line has the header (strings). Fields are separated by a single comma.
Here is an example:
lat,lon,tripID,elapsedTime_sec 35.157885,136.926141,1,0.0 35.15775299999999,136.92623033333334,1,8.082 35.15777023076923,136.92619484615386,1,10.118 35.15771669230769,136.92623269230768,1,12.017 35.15771669230769,136.92623269230768,1,14.019
Note that first and last few seconds (uniform in 5 to 60 seconds) of test trajectories are omitted (but train trajectories are not omitted). The task is hence to infer the future and past of test trajectories and predict the goal and start.
A single txt file of locations of the pre-defined goals.
The first line has the header (strings). Fields are separated by a single comma.
Each txt file describes trips (trajectories).
The first line has the header (strings). Fields are separated by a single comma.
Here is an example of train_trip_info.csv:
tripID,duration,startLocID,destLocID,age,gender 1,7.696116666666667,3,4,26,m 5,13.35055,5,1,22,m 10,2.3670333333333335,4,6,19,f 11,3.53645,2,0,16,f 13,3.07525,4,6,21,m 15,4.80535,5,1,26,m
Note that test_trip_info.csv has only three fields: tripID, age, and gender. Here is an example:
tripID,age,gender 2,50,f 3,20,m 6,7,m 9,23,m
The winner will recieve the following prizes;
Note that, to recieve the prizes, the winners must attend the workshop below.
In addition, the first winner will recieve
Note that
This challenge is governed by the general ChaLearn contest rules.
Start: April 1, 2018, midnight
Description: Development phase: tune your models and submit prediction results, trained model, or untrained model.
Start: Oct. 1, 2018, midnight
Description: Final phase (no submission, your last submission from the previous phase is automatically forwarded).
Oct. 1, 2018, 11:59 p.m.
You must be logged in to participate in competitions.
Sign In