# Latest **URL:** https://forum.menlo.ai/latest.md [Latest](https://forum.menlo.ai/latest.md) · [Categories](https://forum.menlo.ai/categories.md) · [Tags](https://forum.menlo.ai/tags.md) --- ## [Welcome to Menlo Community Forum! 👋](https://forum.menlo.ai/t/welcome-to-menlo-community-forum/5)
**Author:** [@system](https://forum.menlo.ai/u/system)\ **Replies:** 0\ **Last updated:** [March 24, 2026, 7:35am UTC](https://forum.menlo.ai/t/welcome-to-menlo-community-forum/5 "2026-03-24T07:35:10Z")
We are so glad you joined us. Menlo Community Forum This is the forum for everyone building with Asimov and the Menlo stack. Here are some things you can do to get started: :speaking\_head: Introduce yourself by addi… --- ## [A new arm is free. A new gripper is not](https://forum.menlo.ai/t/a-new-arm-is-free-a-new-gripper-is-not/68)
**Author:** [@Ishneet\_Singh](https://forum.menlo.ai/u/Ishneet_Singh)\ **Replies:** 1\ **Last updated:** [September 22, 2026, 1:47am UTC](https://forum.menlo.ai/t/a-new-arm-is-free-a-new-gripper-is-not/68 "2026-09-22T01:47:07Z")
TL;DR: I ran a frozen \`pi05\_libero\` on arms it was never trained on. UR5e and the I2RT YAM both go 20/20, same as the Panda it was fine-tuned on, as long as the wrist camera sits in the same place relative to the fingers… --- ## [Physical AI data collection: Where are today's biggest bottlenecks?](https://forum.menlo.ai/t/physical-ai-data-collection-where-are-todays-biggest-bottlenecks/64)
**Author:** [@vro2](https://forum.menlo.ai/u/vro2)\ **Replies:** 6\ **Last updated:** [August 7, 2026, 6:22am UTC](https://forum.menlo.ai/t/physical-ai-data-collection-where-are-todays-biggest-bottlenecks/64 "2026-08-07T06:22:34Z")
Hi everyone! I’m Olga, founder of VR Oxygen. We built infrastructure for distributed spatial tasks and are now exploring how it can support Physical AI and robotics data collection. I’m hoping to learn from teams alread… --- ## [Why VLAs are so viewpoint brittle?](https://forum.menlo.ai/t/why-vlas-are-so-viewpoint-brittle/66)
**Author:** [@Ishneet\_Singh](https://forum.menlo.ai/u/Ishneet_Singh)\ **Replies:** 3\ **Last updated:** [August 5, 2026, 8:28am UTC](https://forum.menlo.ai/t/why-vlas-are-so-viewpoint-brittle/66 "2026-08-05T08:28:41Z")
TL;DR: a tripod nudge can push the policy much further out of distribution than a completely new object layout. Fixing this likely requires multi-view training, viewpoint augmentation, or representations explicitly encou… --- ## [This week's experiment update: Kepler v0.2](https://forum.menlo.ai/t/this-weeks-experiment-update-kepler-v0-2/63)
**Author:** [@linus](https://forum.menlo.ai/u/linus)\ **Replies:** 0\ **Last updated:** [August 4, 2026, 10:06am UTC](https://forum.menlo.ai/t/this-weeks-experiment-update-kepler-v0-2/63 "2026-08-04T10:06:27Z")
One of our goals with Menlo is to share our research process, including experiments that don’t go as expected. This week, we ran follow-up experiments on Kepler v0.1 to understand why our pretrained encoder wasn’t impro… --- ## [Why Didn't a Better Encoder Produce a Better Robot Policy?](https://forum.menlo.ai/t/why-didnt-a-better-encoder-produce-a-better-robot-policy/62)
**Author:** [@linus](https://forum.menlo.ai/u/linus)\ **Replies:** 0\ **Last updated:** [August 4, 2026, 6:04am UTC](https://forum.menlo.ai/t/why-didnt-a-better-encoder-produce-a-better-robot-policy/62 "2026-08-04T06:04:09Z")
Last week we shared that our pretrained encoder learned good representations but didn’t improve downstream policy performance. Our hypothesis was that the issue was the training objective rather than the architecture, so… --- ## [This week's Kepler v0.2 experiment updates](https://forum.menlo.ai/t/this-weeks-kepler-v0-2-experiment-updates/56)
**Author:** [@linus](https://forum.menlo.ai/u/linus)\ **Replies:** 0\ **Last updated:** [July 27, 2026, 6:02am UTC](https://forum.menlo.ai/t/this-weeks-kepler-v0-2-experiment-updates/56 "2026-07-27T06:02:09Z")
One of our goals with Menlo is to share our research process, including experiments that don’t go as expected. This week, we ran follow-up experiments on Kepler v0.1 to understand why our pretrained encoder wasn’t impro… --- ## [Asimov 1 - #002 Take it first step in the world!](https://forum.menlo.ai/t/asimov-1-002-take-it-first-step-in-the-world/50)
**Author:** [@menlo](https://forum.menlo.ai/u/menlo)\ **Replies:** 0\ **Last updated:** [July 26, 2026, 7:56am UTC](https://forum.menlo.ai/t/asimov-1-002-take-it-first-step-in-the-world/50 "2026-07-26T07:56:49Z")
So some great new, Asimov 1 - 002 is the third robot ever from Menlo Research that takes its first step in the world. Freshly assembled and roll out from our manufacturer assembly line, takes a look below! We are expe… --- ## [How to train a task with Reinforcement learning](https://forum.menlo.ai/t/how-to-train-a-task-with-reinforcement-learning/49)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 16, 2026, 2:01am UTC](https://forum.menlo.ai/t/how-to-train-a-task-with-reinforcement-learning/49 "2026-07-16T02:01:42Z")
Before learning how to train a reinforcement learning policy for robot control, we first need to understand which types of tasks are suitable for RL. For systems that are easy to model and predict, model-based methods a… --- ## [Why we chose Reinforcement Learning to train our locomotion policy](https://forum.menlo.ai/t/why-we-chose-reinforcement-learning-to-train-our-locomotion-policy/48)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 16, 2026, 1:58am UTC](https://forum.menlo.ai/t/why-we-chose-reinforcement-learning-to-train-our-locomotion-policy/48 "2026-07-16T01:58:23Z")
How do Asimov training tasks? What’s the scalable ways to do it? Let’s first understand what are the approaches out there to enable a robot to perform desired tasks and behaviors and the difference between them. In rob… --- ## [Asimov 1 Manual: Request for feedback](https://forum.menlo.ai/t/asimov-1-manual-request-for-feedback/27)
**Author:** [@LazyYuuki\_Menlo](https://forum.menlo.ai/u/LazyYuuki_Menlo)\ **Replies:** 1\ **Last updated:** [July 11, 2026, 2:03pm UTC](https://forum.menlo.ai/t/asimov-1-manual-request-for-feedback/27 "2026-07-11T14:03:51Z")
I am currently working on the rework of the Asimov 1 Manual to make it much clearer and easier to follow along. It is currently a WIP, but I would like to gather more feedback on wording and organization of the content.… --- ## [Work in progress](https://forum.menlo.ai/t/work-in-progress/30)
**Author:** [@Granny](https://forum.menlo.ai/u/Granny)\ **Replies:** 6\ **Last updated:** [July 10, 2026, 11:17am UTC](https://forum.menlo.ai/t/work-in-progress/30 "2026-07-10T11:17:14Z")
Quick summary of my Cognitive stack: The robot has a Raspberry Pi 5 in the head. For now, the Pi is mainly the eyes, ears, speaker, and pipe: camera stream, microphone input, audio output, and basic voice I/O. The heavi… --- ## [Tutorial 7: Motion prior approach to make Asimov walk naturally (Or walk while dancing!)](https://forum.menlo.ai/t/tutorial-7-motion-prior-approach-to-make-asimov-walk-naturally-or-walk-while-dancing/45)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 10:07am UTC](https://forum.menlo.ai/t/tutorial-7-motion-prior-approach-to-make-asimov-walk-naturally-or-walk-while-dancing/45 "2026-07-09T10:07:55Z")
The previous locomotion approach worked well in terms of sim-to-real transfer: the policy could track velocity and handle pushes robustly. However, the gait was still slightly limping and conservative, and further improv… --- ## [Tutorial 6: Rewards Engineering (Define what should Robot learn)](https://forum.menlo.ai/t/tutorial-6-rewards-engineering-define-what-should-robot-learn/44)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 10:06am UTC](https://forum.menlo.ai/t/tutorial-6-rewards-engineering-define-what-should-robot-learn/44 "2026-07-09T10:06:40Z")
In locomotion RL, rewards define what behavior the policy should learn. Instead of directly telling the robot how to move each joint, we define reward terms that encourage desired behaviors and penalize undesirable ones… --- ## [Tutorial 5: Termination](https://forum.menlo.ai/t/tutorial-5-termination/43)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 10:05am UTC](https://forum.menlo.ai/t/tutorial-5-termination/43 "2026-07-09T10:05:03Z")
Termination is used to stop an episode when the robot enters an invalid or unhelpful state, so the policy does not keep learning from messy failure motions. Here we use two termination conditions: fell over termination … --- ## [Tutorial 4: Domain Randomization (How to make the policy robust)](https://forum.menlo.ai/t/tutorial-4-domain-randomization-how-to-make-the-policy-robust/42)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 10:04am UTC](https://forum.menlo.ai/t/tutorial-4-domain-randomization-how-to-make-the-policy-robust/42 "2026-07-09T10:04:17Z")
There are two main ways to reduce the sim-to-real gap. The first is accurate system identification, where we try to model the real robot as precisely as possible in simulation. However, real systems often contain complex… --- ## [Tutorial 3: Actuator modeling](https://forum.menlo.ai/t/tutorial-3-actuator-modeling/41)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 10:01am UTC](https://forum.menlo.ai/t/tutorial-3-actuator-modeling/41 "2026-07-09T10:01:54Z")
Kp and Kd setting In position control, each actuator is commanded to track a desired joint position: τ = Kp(q\_target - q) - Kd q\_dot A higher Kp makes the actuator track the target position more aggressively, improving… --- ## [Tutorial 2: URDF/XML modeling (Understand robot description)](https://forum.menlo.ai/t/tutorial-2-urdf-xml-modeling-understand-robot-description/35)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 9:01am UTC](https://forum.menlo.ai/t/tutorial-2-urdf-xml-modeling-understand-robot-description/35 "2026-07-09T09:01:37Z")
1. how to read/visualize xml An XML file is a structured description of the simulated robot and its environment. In asimov\_1.xml, it defines the simulation environment, robot links and kinematic structure, joints, collis… --- ## [Tutorial 1: Quick Deployment (if you just want to launch the training](https://forum.menlo.ai/t/tutorial-1-quick-deployment-if-you-just-want-to-launch-the-training/34)
**Author:** [@Ariel\_Hu](https://forum.menlo.ai/u/Ariel_Hu)\ **Replies:** 0\ **Last updated:** [July 9, 2026, 8:33am UTC](https://forum.menlo.ai/t/tutorial-1-quick-deployment-if-you-just-want-to-launch-the-training/34 "2026-07-09T08:33:19Z")
Requirements Our training code is based on mjlab, which uses MuJoCo Warp, a GPU accelerated backend optimized for NVIDIA GPUs. Locomotion training, especially with on policy algorithms such as PPO, requires many parallel… --- ## [Sim2Real/Real2Sim Kp Kd](https://forum.menlo.ai/t/sim2real-real2sim-kp-kd/28)
**Author:** [@Ishneet\_Singh](https://forum.menlo.ai/u/Ishneet_Singh)\ **Replies:** 0\ **Last updated:** [July 7, 2026, 4:48am UTC](https://forum.menlo.ai/t/sim2real-real2sim-kp-kd/28 "2026-07-07T04:48:08Z")
Is there any specific decision point or difficulties when tuning Kp / Kd to Simulating or Sim2Real Asimov? i think the main thing we realised for kp kd tuning for sim and real is that its treated as two diff systems to … --- ## [Safety is really important if we are aiming for million of Asimov in the wild!](https://forum.menlo.ai/t/safety-is-really-important-if-we-are-aiming-for-million-of-asimov-in-the-wild/26)
**Author:** [@LazyYuuki\_Menlo](https://forum.menlo.ai/u/LazyYuuki_Menlo)\ **Replies:** 0\ **Last updated:** [July 3, 2026, 1:00pm UTC](https://forum.menlo.ai/t/safety-is-really-important-if-we-are-aiming-for-million-of-asimov-in-the-wild/26 "2026-07-03T13:00:24Z")
This video is a great reminder that we need to own the hardware and software of the next robotic wave. If one day, there is going to be million of Asimov in the wild, we have to collectively make sure that each and every… --- ## [Manual.asimov.inc is live!](https://forum.menlo.ai/t/manual-asimov-inc-is-live/24)
**Author:** [@menlo](https://forum.menlo.ai/u/menlo)\ **Replies:** 0\ **Last updated:** [April 6, 2026, 2:47am UTC](https://forum.menlo.ai/t/manual-asimov-inc-is-live/24 "2026-04-06T02:47:54Z")
We’re releasing a manual for building a humanoid robot. It currently covers Asimov v0 and Asimov v1. Still improving it. Expect blank pages and updates. --- ## [Singapore AI & Robotics Showcase: April](https://forum.menlo.ai/t/singapore-ai-robotics-showcase-april/17)
**Author:** [@menlo](https://forum.menlo.ai/u/menlo)\ **Replies:** 0\ **Last updated:** [March 25, 2026, 8:46am UTC](https://forum.menlo.ai/t/singapore-ai-robotics-showcase-april/17 "2026-03-25T08:46:13Z")
Let’s meet in Singapore! Tuesday, April 14 7pm-8.30pm (doors open 6.30pm) ​​​​National Library Board, Level 7 - Launch Programme Room RSVP Required: