Skip to content

Commit 87c49b5

Browse files
committed
.
1 parent 5768df6 commit 87c49b5

File tree

1 file changed

+8
-13
lines changed

1 file changed

+8
-13
lines changed

public/OpenGO1/index.html

Lines changed: 8 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
<head>
22
<link rel="icon" type="image/png" href="./colosseo.png">
3-
<title>AgiBot World Colosseo | OpenDriveLab</title>
3+
<title>OpenGO1 | OpenDriveLab</title>
44
<meta charset="UTF-8">
55
<meta name="viewport" content="width=device-width, initial-scale=1.0">
66
<meta name="keywords" content="AgiBot World, OpenDriveLab, Robotics, Embodied AI, Autonomous Driving">
@@ -83,15 +83,10 @@ <h2 class="text-white w-full max-w-landing mb-12">
8383
<img loading="lazy" src="/assets/brand/hku.png" class="h-5 laptop:h-10 select-none"/>
8484
<img loading="lazy" src="/assets/brand/agibot_white.png" class="h-4 laptop:h-8 select-none"/>
8585
<img loading="lazy" src="/assets/brand/shanghai_innovation_institute_white.png" class="h-4 laptop:h-8 select-none"/>
86-
<img loading="lazy" src="/assets/brand/shanghai_ai_lab_white_.png" class="h-4 laptop:h-8 select-none"/>
8786
</div>
8887

8988
<p class="text-white w-full max-w-landing mb-6 select-none">
9089
<a class="text-white underline hover:text-o-blue" href="https://arxiv.org/abs/2503.06669" target="_blank">Technical Report</a>
91-
<span class="select-none text-white">&nbsp;|&nbsp;</span>
92-
<a class="text-white underline hover:text-o-blue" href="https://youtu.be/foTDYU1lIPE" target="_blank">Video</a>
93-
<span class="select-none text-white">&nbsp;|&nbsp;</span>
94-
<a class="text-white underline hover:text-o-blue" href="https://opendrivelab.github.io/content/poster_AgiBot_World_Colosseo.pdf" target="_blank">Poster</a>
9590
</p>
9691

9792
<p class="text-white w-full max-w-landing mb-12 select-none">
@@ -162,7 +157,7 @@ <h2 class="text-white w-full max-w-landing mb-12">
162157

163158

164159
<p class="text-white w-full max-w-wide leading-loose">
165-
We are open-sourcing our generalist robotic foundation model, GO-1. Beyond the dataset and model innovations we shared in our <a href="https://opendrivelab.com/AgiBot-World/" class="text-white underline hover:text-o-blue">previous blog</a>, this time wed like to talk about the bitter lessons we learned along the way.
160+
We are open-sourcing our generalist robotic foundation model, GO-1. Beyond the dataset and model innovations we shared in our <a href="https://opendrivelab.com/AgiBot-World/" class="text-white underline hover:text-o-blue">previous blog</a>, this time we'd like to talk about the bitter lessons we learned along the way.
166161
</p>
167162

168163
<p class="text-white w-full max-w-wide leading-loose">
@@ -174,7 +169,7 @@ <h2 class="text-white w-full max-w-landing mb-12">
174169
</p>
175170

176171
<p class="text-white w-full max-w-wide leading-loose">
177-
Even if a team manages to squeeze out a 5% or 10% gain in model accuracy, the entire system can still fail if any part of the pipeline is broken. This is the bucket effect. A wrong coordinate frame, inconsistencies between data collection and execution, or even a minor hardware failure can cause the robots actions to collapse completely. When scaling up data collection to more than a hundred robots, even more factors need to be considered.
172+
Even if a team manages to squeeze out a 5% or 10% gain in model accuracy, the entire system can still fail if any part of the pipeline is broken. This is the bucket effect. A wrong coordinate frame, inconsistencies between data collection and execution, or even a minor hardware failure can cause the robot's actions to collapse completely. When scaling up data collection to more than a hundred robots, even more factors need to be considered.
178173
</p>
179174

180175
<p class="text-white w-full max-w-wide leading-loose">
@@ -219,11 +214,11 @@ <h4 class="text-white w-full max-w-wide leading-loose ">
219214
</p>
220215

221216
<p class="text-white w-full max-w-wide leading-loose">
222-
Of course, robots arent perfect. Because of differences in their built-in controllers and software, the arm might show grasping errors anywhere from 1 mm to 10 mm. Youll notice this if the robot drops the screw or if the screw ends up off-center in the gripper. These errors highlight the gap between what you demonstrated and what the robot actually executes. And if you train a VLA model with data that has these kinds of errors, it will likely struggle to perform the task, even if the model itself is strong and well trained.
217+
Of course, robots aren't perfect. Because of differences in their built-in controllers and software, the arm might show grasping errors anywhere from 1 mm to 10 mm. You'll notice this if the robot drops the screw or if the screw ends up off-center in the gripper. These errors highlight the gap between what you demonstrated and what the robot actually executes. And if you train a VLA model with data that has these kinds of errors, it will likely struggle to perform the task, even if the model itself is strong and well trained.
223218
</p>
224219

225220
<p class="text-white w-full max-w-wide leading-loose">
226-
Now, things get even trickier when you scale up. Imagine collecting data with dozens or even hundreds of robots at the same time, like we do in our data collection factory. In this case, its not enough for data to work on the robot it was collected from, you also need it to work across different robots. That way, all the data can be treated as one big, unified dataset, rather than being tied to a specific machine. This cross-robot consistency not only boosts scalability but also makes it possible to evaluate models on any robot in the fleet.
221+
Now, things get even trickier when you scale up. Imagine collecting data with dozens or even hundreds of robots at the same time, like we do in our data collection factory. In this case, it's not enough for data to work on the robot it was collected from, you also need it to work across different robots. That way, all the data can be treated as one big, unified dataset, rather than being tied to a specific machine. This cross-robot consistency not only boosts scalability but also makes it possible to evaluate models on any robot in the fleet.
227222
</p>
228223

229224
<h4 class="text-white w-full max-w-wide leading-loose ">
@@ -289,7 +284,7 @@ <h4 class="text-white w-full max-w-wide leading-loose ">
289284
</h4>
290285

291286
<p class="text-white w-full max-w-wide leading-loose">
292-
The action space defines the coordinates in which robots operate. Traditional robot controllers often use the end-effector (EEF) pose, measured relative to the robot’s chest or base and compared to the previous frame. In the VLA era, however, a more direct approach is to control the arm motors through their joint angles. There are also multiple ways to design the learning objectives for a model. For example, predicting actions relative to the last frame, or relative to the first frame in an action chunk (as in pi0). From our experiments, we’ve seen that strong models can easily adapt to different action spaces, even for dexterous manipulation tasks like cloth folding. In fact, they can learn effectively even when the pre-training and fine-tuning stages use different action spaces. To keep things simple for users, we choose to adopt absolute joint space in our open-source model. The key takeaway here is that the robot must execute the actions predicted by the model correctly. Since some robot controllers operate in different coordinate systems, a coordinate transformation may be needed to ensure everything lines up properly.
287+
The action space defines the coordinates in which robots operate. Traditional robot controllers often use the end-effector (EEF) pose, measured relative to the robot's chest or base and compared to the previous frame. In the VLA era, however, a more direct approach is to control the arm motors through their joint angles. There are also multiple ways to design the learning objectives for a model. For example, predicting actions relative to the last frame, or relative to the first frame in an action chunk (as in pi0). From our experiments, we've seen that strong models can easily adapt to different action spaces, even for dexterous manipulation tasks like cloth folding. In fact, they can learn effectively even when the pre-training and fine-tuning stages use different action spaces. To keep things simple for users, we choose to adopt absolute joint space in our open-source model. The key takeaway here is that the robot must execute the actions predicted by the model correctly. Since some robot controllers operate in different coordinate systems, a coordinate transformation may be needed to ensure everything lines up properly.
293288
</p>
294289

295290
<h4 class="text-white w-full max-w-wide leading-loose ">
@@ -327,11 +322,11 @@ <h4 class="text-white w-full max-w-wide leading-loose ">
327322
</h4>
328323

329324
<p class="text-white w-full max-w-wide leading-loose">
330-
Now that we have a trained VLA model, its time for evaluation. But dont rush straight into deploying it on a real robot. A good first step is to run an open-loop test, which validates whether the model has properly fit the fine-tuning data. If the model fails here, it usually points to issues in your pipeline, such as problems with the dataset, dataloader, or training process. Fix them before moving forward.
325+
Now that we have a trained VLA model, it's time for evaluation. But don't rush straight into deploying it on a real robot. A good first step is to run an open-loop test, which validates whether the model has properly fit the fine-tuning data. If the model fails here, it usually points to issues in your pipeline, such as problems with the dataset, dataloader, or training process. Fix them before moving forward.
331326
</p>
332327

333328
<p class="text-white w-full max-w-wide leading-loose">
334-
Once the model passes the open-loop test, you can deploy it to the robot and begin real-world testing. As an extra precaution, its also a good idea to first replay previously collected data on the robot to confirm that all hardware is functioning as expected.
329+
Once the model passes the open-loop test, you can deploy it to the robot and begin real-world testing. As an extra precaution, it's also a good idea to first replay previously collected data on the robot to confirm that all hardware is functioning as expected.
335330
</p>
336331

337332
<h4 class="text-white w-full max-w-wide leading-loose ">

0 commit comments

Comments
 (0)