Method for performing simultaneous localization and mapping and device using same
Provided is an accelerator provided in an electronic device and configured to perform simultaneous localization and mapping (SLAM), the accelerator including a factor graph database, a memory, and a back-end processor, wherein the back-end processor is configured to receive a first piece of data corresponding to map points and camera positions from the factor graph database, convert the received first piece of data into a matrix for the map points and a matrix for the camera positions, store, in the memory, results obtained by performing an optimization calculation on the matrix for the map points and a matrix for at least one camera position, among the camera positions, corresponding to the map points, and obtain a second piece of data optimized with respect to the first piece of data based on the results stored in the memory.
1. An accelerator provided in an electronic device and configured to perform simultaneous localization and mapping (SLAM), the accelerator comprising:
a factor graph database;
a memory; and
a back-end processor configured to:
receive a first piece of data corresponding to map points and camera positions from the factor graph database;
convert the received first piece of data into a matrix for the map points and a matrix for the camera positions;
divide the matrix for the map points and the matrix for at least one camera position corresponding to the map points into a plurality of sub-tracks, the plurality of sub-tracks having a track length set based on a number of camera positions on which the back-end processor is configured to simultaneously perform calculations;
store, in the memory, results obtained by performing an optimization calculation on the matrix for the map points and a matrix for at least one camera position, among the camera positions, corresponding to the map points; and
obtain a second piece of data optimized with respect to the first piece of data based on the results stored in the memory,
wherein the back-end processor is further configured to perform an optimization calculation on matrices corresponding to the plurality of sub-tracks based on a plurality of shift registers, the plurality of shift registers corresponding to the plurality of sub-tracks, and
wherein pieces of data processed through the plurality of shift registers are sequentially input to a timing controller and the timing controller is configured to simultaneously transmit the pieces of data.
2. The accelerator of claim 1 , wherein the back-end processor is further configured to perform an optimization calculation on matrices corresponding to the plurality of sub-tracks based on a shift register.
3. The accelerator of claim 1 , wherein the back-end processor is further configured to:
store, in the memory, a first result obtained by performing an optimization calculation on a matrix for a first map point corresponding to a first sub-track of the plurality of sub-tracks, and on a matrix for at least one camera position corresponding to the first map point; and
obtain a second result by performing an optimization calculation on a matrix for a second map point corresponding to a second sub-track following the first sub-track in the plurality of sub-tracks, and on a matrix for at least one camera position corresponding to the second map point.
4. The accelerator of claim 3 , wherein the back-end processor is further configured to store the obtained second result in the memory in which the first result is stored.
5. The accelerator of claim 3 , wherein based on obtaining the second result, the back-end processor is further configured to reuse a matrix corresponding to at least a portion of the second sub-track, the portion corresponding to the first sub-track.
6. The accelerator of claim 1 , wherein based on performing the optimization calculation, the back-end processor is further configured to perform a matrix calculation and a vector calculation in parallel.
7. A method of performing simultaneous localization and mapping (SLAM) by an accelerator provided in an electronic device, the method comprising:
receiving a first piece of data corresponding to map points and camera positions from a factor graph database;
converting the received first piece of data into a matrix for the map points and a matrix for the camera positions;
dividing the matrix for the map points and the matrix for the at least one camera position corresponding to the map points into a plurality of sub-tracks, the plurality of sub-tracks having a track length which is set based on a number camera positions simultaneously calculated;
storing, in a memory, results obtained by performing an optimization calculation on the matrix for the map points and a matrix for at least one camera position, among the camera positions, corresponding to the map points; and
obtaining a second piece of data optimized with respect to the first piece of data based on the results stored in the memory,
wherein the method further comprises performing an optimization calculation on matrices corresponding to the plurality of sub-tracks based on a plurality of shift registers, the plurality of shift registers corresponding to the plurality of sub-tracks, and
wherein pieces of data processed through the plurality of shift registers are sequentially input to a timing controller and the timing controller is configured to simultaneously transmit the pieces of data.
8. The method of claim 7 , further comprising performing an optimization calculation on matrices corresponding to the plurality of sub-tracks based on a shift register.
9. The method of claim 7 , further comprising:
storing, in the memory, a first result obtained by performing an optimization calculation on a matrix for a first map point corresponding to a first sub-track of the plurality of sub-tracks, and on a matrix for a camera position corresponding to the first map point; and
obtaining a second result by performing an optimization calculation on a matrix for a second map point corresponding to a second sub-track following the first sub-track in the plurality of sub-tracks, and on a matrix for a camera position corresponding to the second map point.
10. The method of claim 9 , further comprising storing the obtained second result in the memory in which the first result is stored.
11. The method of claim 9 , wherein based on obtaining the second result, the method further comprises reusing a matrix corresponding to at least a portion of the second sub-track, the portion corresponding to the first sub-track.
12. An electronic device configured to perform simultaneous localization and mapping (SLAM), the electronic device comprising:
a sensor;
a factor graph database;
a memory; and a processor configured to control the sensor and the memory,
wherein the processor is further configured to:
perform a key point extraction calculation and a coordinate calculation on sensing data received from the sensor, and transmit results of the key point extraction calculation and the coordinate calculation to the factor graph database;
obtain a first piece of data corresponding to map points and camera positions from the factor graph database;
convert the obtained first piece of data into a matrix for the map points and a matrix for the camera positions;
divide the matrix for the map points and the matrix for at least one camera position corresponding to the map points into a plurality of sub-tracks, the plurality of sub-tracks having a track length set based on a number of camera positions on which the processor is configured to simultaneously perform calculations;
store, in the memory, results obtained by performing an optimization calculation on the matrix for the map points and a matrix for at least one camera position, among the camera positions, corresponding to the map points; and
based on the results stored in the memory,
wherein the processor is further configured to perform an optimization calculation on matrices corresponding to the plurality of sub-tracks based on a plurality of shift registers, the plurality of shift registers corresponding to the plurality of sub-tracks, and
wherein pieces of data processed through the plurality of shift registers are sequentially input to a timing controller and the timing controller is configured to simultaneously transmit the pieces of data.
13. The electronic device of claim 12 , wherein the processor is further configured to: store, in the memory, a first result obtained by performing an optimization calculation on a matrix for a first map point corresponding to a first sub-track of the plurality of sub-tracks, and on a matrix for a camera position corresponding to the first map point; and obtain a second result by performing an optimization calculation on a matrix for a second map point corresponding to a second sub-track following the first sub-track in the plurality of sub-tracks, and on a matrix for a camera position corresponding to the second map point.
14. The electronic device of claim 13 , wherein based on obtaining the second result, the processor is further configured to reuse a matrix in at least a portion of the second sub-track, the portion corresponding to the first sub-track.