WEBVTT 00:00.000 --> 00:11.000 Hi everyone, so this is a presentation about modernizing 00:11.000 --> 00:12.000 Rostos skills. 00:12.000 --> 00:14.000 I'm here in Focicale. 00:14.000 --> 00:18.000 I'm the sexual well-engineered at University College London. 00:18.000 --> 00:21.000 I'm presenting this go on behalf of the Advanced Society 00:21.000 --> 00:24.000 Competition and the Geomatic Engineering teams. 00:24.000 --> 00:29.000 So mainly the challenge is how we teach all these 00:29.000 --> 00:34.000 like computer and hardware skills to students. 00:34.000 --> 00:38.000 We need to manage lots of things, 00:38.000 --> 00:42.000 the server, the virtual machines, the robot sensors, 00:42.000 --> 00:46.000 and the way we are trying to do this is by using cloud computing. 00:46.000 --> 00:50.000 And the reason is because participants 00:50.000 --> 00:53.000 or persons have different skills and levels, 00:53.000 --> 00:56.000 so we need to handle all these kind of dependencies 00:56.000 --> 01:01.000 of making sure they strive for where to start the developing. 01:01.000 --> 01:06.000 So what we are trying to do is orchestrating a few of the infrastructure. 01:06.000 --> 01:13.000 Like a bare metal server that runs all the packages 01:13.000 --> 01:17.000 that collect the sensor data, control the robots, 01:17.000 --> 01:22.000 and we use a signal router to communicate through 01:23.000 --> 01:27.000 Rostos middle-guests through these virtual machines. 01:27.000 --> 01:30.000 So this metal machine is making easier for users to start 01:30.000 --> 01:34.000 connecting to them via SSH and start playing with 01:34.000 --> 01:37.000 Rostos packages, Rostos Playbacks. 01:37.000 --> 01:41.000 They really start working on those kind of life fun stuff. 01:41.000 --> 01:43.000 And not really struggling, 01:43.000 --> 01:48.000 I need to install my Python version 3.12, 01:48.000 --> 01:51.000 3.11, 3.12 and playing with all these flavors of Rostos. 01:51.000 --> 01:54.000 And everything, which is also fun for me, 01:54.000 --> 01:56.000 but not maybe for everyone. 01:56.000 --> 02:00.000 So that's why we are kind of playing with this infrastructure. 02:00.000 --> 02:04.000 So this is like for you kind of like a demos. 02:04.000 --> 02:07.000 We have been doing this kind of workflow 02:07.000 --> 02:12.000 where we develop the Docker containers with all the Rostos dependencies 02:12.000 --> 02:16.000 with them, create the GitHub container registry, 02:16.000 --> 02:20.000 which is make it easier to make it portable. 02:20.000 --> 02:25.000 And then that way we pull that image into the virtual machine 02:25.000 --> 02:29.000 and that's more easier for users to start 02:29.000 --> 02:31.000 just playing with the data. 02:31.000 --> 02:34.000 So that's what we are doing now. 02:34.000 --> 02:37.000 We have done this virtual and virtual 02:37.000 --> 02:39.000 and virtual metal communications. 02:39.000 --> 02:45.000 We are this and uploading Rostos DS. 02:45.000 --> 02:48.000 Also the other thing, we are playing around 02:48.000 --> 02:53.000 is how we interact with GPU, server capacities. 02:53.000 --> 02:57.000 So for example, the ISAXINA and ISA lab. 02:57.000 --> 02:59.000 So we have been running this into the virtual machine. 02:59.000 --> 03:02.000 Just the users, just connect to SSH. 03:02.000 --> 03:05.000 And then start playing with the interaction 03:05.000 --> 03:08.000 and all the kind of the models. 03:08.000 --> 03:12.000 So lots of things that we want to do 03:12.000 --> 03:17.000 in the future, like we more of the multi-modal sensing 03:17.000 --> 03:21.000 and going from the low level to the high level attention. 03:21.000 --> 03:25.000 So there's a lot of things that depends on the sensor we focus on. 03:25.000 --> 03:31.000 And of course, we are kind of also developing, sorry about this. 03:31.000 --> 03:35.000 But we are working on the, I don't know what's happening. 03:35.000 --> 03:41.000 But one thing we are working on is developing visual language models. 03:42.000 --> 03:47.000 And action models are also the visual language navigation models. 03:47.000 --> 03:49.000 Sorry about this. 03:49.000 --> 03:52.000 But that's what we are trying to do. 03:52.000 --> 03:54.000 That's kind of hit a wall. 03:54.000 --> 03:59.000 The only thing we are working on is on healthcare 03:59.000 --> 04:04.000 and robotics, medical robotics applications. 04:04.000 --> 04:09.000 So we've been organizing workshops. 04:09.000 --> 04:12.000 The handling medical symposium. 04:12.000 --> 04:16.000 So last year, we have like 50 participants, 04:16.000 --> 04:20.000 10 speakers, these posters, and Ben was all out. 04:20.000 --> 04:23.000 And we are organizing a new event. 04:23.000 --> 04:27.000 If anyone is interested in around these topics, you just talk to me. 04:27.000 --> 04:30.000 And lastly, we have like a few takeaways. 04:30.000 --> 04:34.000 What we want to do is lower in the barriers to advanced robotics 04:34.000 --> 04:35.000 as a scale. 04:35.000 --> 04:38.000 We have a proven and scale model for collaboration. 04:38.000 --> 04:42.000 We have a test bed for research training and innovation. 04:42.000 --> 04:46.000 And of course, if anyone is keen to work with us, 04:46.000 --> 04:51.000 or collaborate, please talk to me or Marlon who is also in the audience. 04:51.000 --> 04:53.000 Yeah, thank you. 04:53.000 --> 04:56.000 And to all the colleagues from UCL and the panelists. 04:56.000 --> 04:59.000 And yeah, we have everything on GitHub as well on the website as well. 04:59.000 --> 05:01.000 So yeah, thank you very much. 05:01.000 --> 05:07.000 Thank you so much Miguel.