Skip to main content
Machina extends Adapt-1 Preview with continuous-control sequence learning. Adapt-1 proposes numeric commands through your permitted control interface and uses measured outcomes to revise sequences across attempts. The resulting executions remain available for later use. Your application measures observations and executes the commands through its actuator mapping. Machina uses Adapt-1’s contextual and temporal learning in this trajectory and procedural-control workflow.

Configure your first control task

Define numeric inputs and controls, then create a trajectory Domain.

Choose an execution pattern

Choose command timing, observation boundaries, and how acquired routines fit together.

​
Control applications

Use Machina to acquire and refine control behavior for robots, machines, and interactive environments. Your integration supplies the controls, observations, and measured outcomes. Define a measurable task and evaluate retained behavior in the environment where your application will use it. The use cases page also covers Adapt-1 workflows that use predictions, policy choices, or retained evidence.

​
What your application sends and receives

A command is one row of control values applied together. The horizon limits the number of rows in a sequence, and control width specifies the values in each row. For example, eight two-coordinate commands contain sixteen scalar outputs. Your executor sets the duration of each command.

​
Learn through execution

1

Acquire a sequence

Request controls, execute them, and return what actually happened. Later attempts can revise retained executions, including a useful prefix followed by a different continuation. Start with sequence acquisition.
2

Test structural changes

Compare acquired executions on repeatable contexts. Test removing spans or changing the order of control changes, then carry the selected execution forward. Use selection, reduction, and ordering.
3

Refine for the current context

Preserve the acquired base and learn bounded corrections from pre-action observations. Use contextual refinement, including optional context normalization and precision settings when corrections vary across operating conditions.
4

Reuse or continue learning

Execute retained behavior with learning stopped, or resume compatible practice. Keep the same Domain identity, base, and executor contract. See retained use and recovery.
Your application invokes the stages through the API as the task requires. A structural stage may retain its input unchanged. Evaluate refinement against the acquired base.

​
Define the executor contract

The executor maps each command to the declared operation. Adapt-1 acquires which commands to issue and how to coordinate them. The execution patterns explain target values, rates, timing, and context-dependent corrections. Mechanisms places acquired procedures and contextual corrections alongside Core’s other forms of learning.

​
Choose when observations can change the controls

During a committed sequence, low-level controllers apply the commands already proposed by Adapt-1. A new Adapt-1 proposal requires another request at an observation boundary. See execution windows for the integration loop.
A frozen contextual policy computes corrections for the current observation using unchanged learned state. Executing the base directly repeats the acquired sequence. See retained use for evaluation and reconnect checks.

​
Connect the right API

The numeric workflow uses POST /domains/{domain_id}/trajectory/{operation} under the Adapt-1 production API. Its configuration and observation bodies are specific to the selected mechanism. Start with Schemas and configuration for authentication, request bodies, and the shared Python helper. The guides use a compact numeric example for the API shapes. Your application supplies the environment hooks.