Communication protocol

The actions and conventions that specify the TAS communication protocol can be found below. They build up on the general part of the communication protocol.

Actions

Action	Specific entries in data		Description
	Client → Server	Server → Client
ping	–	method version	Client checks connection and gets information about the server. method: string Method used for the server version: string Version of the server
reset	mode axes offset limits level_backgr thresh_intens travel_cost_max scenario_name	–	Client resets the server for a new scan. mode: string Currently, only “single” is available axes: list of 4D lists of floats Axes in \(\mathbf{Q}\)-\(E\) space offset: list of floats Offset for the axes limits: list of two 2D lists of floats Limits of the scan area, length same as that of axes level_backgr: float, optional Value for background level thresh_intens: float, optional Value for intensity threshold, larger than level_backgr travel_cost_max: float, optional Maximum value for travel costs (used for normalization), default is 1.0. scenario_name: string, optional Name of scenario, included in name of log files if specified
result	locs counts matrices_ellipses travel_time counting_time travel_cost_grid travel_cost_values	–	Client sends results of measurements (counts) to the server. A discretization of the cost function at the current location locs[-1] can also be provided, if applicable. locs: list of nD points Measurement locations counts: list of pairs Pairs of detector and monitor counts, length same as that of locs matrices_ellipses: list of n x n matrices Matrices encoding an ellipse shaped zone that should be excluded from suggesting new locations in, length same as that of locs travel_time: float Time needed to move axes for counting at locs counting_time: float Time needed to count at locs travel_cost_grid: list of nD points, optional Grid in hyperrectangle given by limits for cost function travel_cost_values: list of floats, optional Cost function values at locations in travel_cost_grid, given if and only if travel_cost_grid is given, length same as that of travel_cost_grid
next_loc	–	loc stop	Client requests the next location to measure at and whether it should stop the experiment at the server. loc: nD list of floats Next location stop: bool Boolean specifying whether the experiment should be stopped
heuris_experi_param	–	level_backgr thresh_intens	Client requests heuristically computed experiment parameters at the server. If the computation is not possible, the server returns None values. level_backgr: float Value for background level thresh_intens: float Value for intensity threshold, larger than level_backgr
problem_locs	locs matrices_ellipses	–	Client informs server about problem locations. locs: list of nD points Problematic locations matrices_ellipses: list of n x n matrices Matrices encoding ellipse shaped zones that should be excluded from suggesting new locations in, length same as that of locs
state_internal	depending on method	depending on method	Client requests the internal state of the approach (see table below) at the server.
stop	–	–	Client stops the experiment at the server.

Internal states

Server method	Data		Description
	Client → Server	Server → Client
GPR	num	grid means stds	num: list of ints or int Number of points for each dimension in grid, length is \(n\) (if list) grid: list of nD points Grid in hyperrectangle given by limits means: list of floats Posterior mean function of log-GP at locations in grid, length same as that of grid stds: list of floats Posterior standard deviation function of log-GP at locations in grid, length same as that of grid

Conventions

• Each experiment starts with reset.

• Each initial measurement point is sent via results.

• The first message with next_loc triggers the initialization of the method at the server. Note that during initialization, which usually takes some time, the server responds with busy = True.

Example workflow

Test connection:

1. Client → Server: ["ARIANE", "", "ping", "{}"]

2. Server → Client: ["ARIANE", "", "ping", "{ success: True, method: 'GPR', version: '0.0.1' }"]

Reset server:

3. Client → Server: ["ARIANE", "", "reset", "{ mode: 'single', axes: [[1, 0, 0, 0], [0, 0, 0, 1]], offset: [0, 1, 1, 0], limits: [[1, 3], [2, 9]] }"]

4. Server → Client: ["ARIANE", "", "reset", "{ success: True }"]

Initialize server:

5. Client → Server: ["ARIANE", "", "result", "{ locs: [[1, 2]], counts: [[30, 100000]], matrices_ellipses: [[[625.0, 0.0], [0.0, 23.5]]] }"]

6. Server → Client: ["ARIANE", "", "result", "{ success: True }"]

7. … (client sends all initialization points)

8. Client → Server: ["ARIANE", "", "result", "{ locs: [[3, 9]], counts: [[12, 100000]], matrices_ellipses: [[[625.0, 0.0], [0.0, 23.5]]] }"]

9. Server → Client: ["ARIANE", "", "result", "{ success: True }"]

10. Client → Server: ["ARIANE", "", "next_loc", "{}"]

11. Server → Client: ["ARIANE", "", "next_loc", "{ success: True, busy: True }"]

12. … (client keeps sending messages with next_loc until the server is not busy anymore)

Main loop:

13. Client → Server: ["ARIANE", "", "next_loc", "{}"]

14. Server → Client: ["ARIANE", "", "next_loc", "{ success: True, loc: [2, 5.5] }"]

15. Client → Server: ["ARIANE", "", "result", "{ locs: [[2, 5.5]], counts: [[170, 100000]], matrices_ellipses: [[[625.0, 0.0], [0.0, 23.5]]] }"]

16. Server → Client: ["ARIANE", "", "result", "{ success: True }"]

17. … (client requests next locations and sends back the results until it stops the server)

Stop server:

18. Client → Server: ["ARIANE", "", "stop", "{}"]

19. Server → Client: ["ARIANE", "", "stop", "{ success: True }"]

This example represents an ideal workflow. The following actions were not considered:

1. heuris_experi_param: to print experiment parameters for the user’s information, as these are computed heuristically if they were not specified during reset.

2. problem_locs: to exclude ellipses around problem locations as potential regions for next measurement points.

3. state_internal: to print or visualize a representation of the server method’s internal state.