Kyros

Integrantes

Inspiration

Kyros was inspired by the idea of turning a robotic arm into a helpful desk assistant, capable of interacting with a human and the environment in a natural way. For the RoBorregos Hack Days challenge, we wanted to build more than just a robot that moves; we wanted to create a simulation where the robot could understand a user request, make a decision, and execute a useful action inside a 3D environment. The challenge encouraged projects focused on autonomy, perception, planning, and control, which became the foundation of our idea.

What it does

Kyros is a simulated robotic desk assistant built in MuJoCo. The robot acts like an intelligent lamp/assistant that can respond to voice commands, interpret the user’s intention with Gemini, and perform physical actions in the simulation. It can move its robotic arm, adjust its lamp direction, and interact with its environment based on the task requested by the user.

The goal of Kyros is to demonstrate how a robot can combine simulation, artificial intelligence, and robotic control to create a more natural human-robot interaction experience.

How we built it

We built Kyros using Python, MuJoCo, and Gemini. MuJoCo was used as the 3D physics simulation environment because it allows us to model robotic movement, joints, lighting, and physical interactions. Python was used to organize the main logic of the project, including the robot controller, voice interaction, command processing, and task execution.

Gemini was integrated as the intelligence layer. The user can give a voice command, the system converts it into text, sends it to Gemini, and then interprets the response to decide what action the robot should perform. We structured the project in modules so the simulation, AI assistant, and robot control could work together in a cleaner way.

Challenges we ran into

One of the biggest challenges was connecting the AI assistant with the robotic simulation in a way that felt smooth and functional. It was not enough for the robot to move; the movement had to match the user’s intention.

We also ran into technical issues with MuJoCo poses, joints, and object positioning. Some parts of the robot appeared in the wrong place or did not move as expected, so we had to adjust the model, joint references, and control values. Another challenge was handling API limits when using Gemini, especially during repeated voice interactions.

Accomplishments that we're proud of

We are proud that we created a functional concept that combines robotics, simulation, voice interaction, and AI decision-making in one project. Even with limited hackathon time, Kyros demonstrates a clear idea: a robot that can understand a human command and translate it into movement inside a simulated environment.

We are also proud of the modular structure of the project, because it gives us a solid base to keep improving the robot after the hackathon. The project is not only a demo; it is a starting point for a more advanced autonomous assistant.

What we learned

We learned how important simulation is for testing robotic ideas before moving to real hardware. We also learned that integrating AI with robotics requires more than just sending prompts; the system needs a clear bridge between language, decisions, and physical actions.

Another key lesson was the importance of defining a realistic scope. In a hackathon, it is better to build a small but working system than to aim for a huge idea that cannot be demonstrated clearly.

What's next for Kyros

The next step for Kyros is to increase its autonomy. We want the robot to perceive more information from the environment, make better decisions, and complete more complex tasks without direct human control.

Future improvements include adding object detection, more precise arm manipulation, smoother motion planning, better voice interaction, and a larger set of commands. Eventually, Kyros could evolve from a simulated desk assistant into a real physical robotic assistant capable of helping with daily tasks in homes, offices, or learning environments.