Chemistry Lab Automation via Constrained Task and Motion Planning

Under Review
^*Indicates Equal Contribution ¹ University of Toronto & Vector Institute ² University of Waterloo ³ Nvidia

Abstract

Chemists need to perform many laborious and time-consuming experiments in the lab to discover and understand the properties of new materials. To support and accelerate this process, we propose a robot framework for manipulation that autonomously performs chemistry experiments. Our framework receives high-level abstract descriptions of chemistry experiments, perceives the lab workspace, and autonomously plans multi-step actions and motions. The robot interacts with a wide range of lab equipment and executes the generated plans. A key component of our method is constrained task and motion planning using PDDLStream solvers. Preventing collisions and spillage is done by introducing a constrained motion planner. Our planning framework can conduct different experiments employing implemented actions and lab tools. We demonstrate the utility of our framework on pouring skills for various materials and two fundamental chemical experiments for materials synthesis: solubility and recrystallization.

Framework for Robot-assisted Chemical Synthesis

We've developed a new framework for robot-assisted chemical synthesis that includes Perception, Task & Motion Planning, and Skills blocks. This framework allows robots to utilize available lab devices (such as sensors and actuators) by adding them to the robot network via ROS. The robot is also equipped with an extra degree of freedom at its end-effector, enabling it to perform constrained motions. Our framework receives the chemical synthesis goal in XDL format, and converts the procedure component into corresponding PDDL goals. The hardware and reagents components then identify the necessary initial conditions for synthesis. Perception is used to detect objects, estimate their positions and contents in the workspace, and monitor task progress. Finally, PDDLStream generates a sequence of actions for the robot to execute.

Constrained Task and Motion Planning

In a chemistry lab, the robot tends to carry beakers that contain liquids inside. Spillage avoidance should be considered in generating robot motions. Our framework integrates constrained motion planner with task and motion planning module and realized spillage-free robot motion.

Constrained Motion Planning Performance of 7 DoF and 8 DoF Robot

The constrained motion planning performance of 7 DoF and 8 DoF robot is evaluated in two scenarios: (1) single step, (2) two steps. In scenario (1), robots find a constrained path with a fixed orientation from initial to final positions that are randomly sampled. Scenario (2) extends the first with an additional intermediate sampled waypoint.

Scenario 1: Single Step

7 Degrees of Freedom

8 Degrees of Freedom

Scenario 2: Two Step

7 Degrees of Freedom

8 Degrees of Freedom

The performance of the 7-DoF and 8-DoF robot arms for the two scenarios are shown in the following table. The results show that the IK and constrained motion planning have higher success rates in 8-DoF compared with the 7-DoF robot.

	Scenario 1 (%)		Scenario 2 (%)
	IK	Plan	IK	Plan
7 DoF	99	84	99	70
8 DoF	100	97	100	84

Pouring skill

Liquid transfer is one of the essential skills used in a chemistry lab. A pouring skill that intermittently rotates the container is implemented. This pouring strategy compensates the delayed feedback from the scale and outperformed PD controller. We present and analyze a set of accurate and efficient pouring skills inspired by human motions.