An Intelligent Robotic System for Perceptive Pancake Batter Stirring and Precise Pouring

TL;DR

This work tackles automated pancake production by integrating perception, stirring, and pouring for viscous, non-Newtonian batter. It leverages force-torque sensing during push maneuvers to jointly estimate batter uniformity, liquid level, and water-flour ratio, and uses these estimates to drive open-loop pouring with stroke-width control via MLP-based speed prediction. Key contributions include a four-motion stirring protocol, a torque-based perception pipeline, and trajectory-planning methods to render arbitrary shapes with consistent line widths and pancake diameters, achieving uniform batter in multiple trials and low estimation errors (e.g., water-flour ratio $\leq 9.6\%$, liquid level $\leq 3.88\%$, line width $\leq 9.6\%$). The system demonstrates practical kitchen applicability by enabling shape-controlled pours and providing a path toward commercial culinary automation, while acknowledging current limits on sharp-edge turns and suggesting future shape-forming improvements. Overall, the paper presents a cohesive, perception-informed robotic framework that realises both functional batter preparation and creative, shape-based pouring.

Abstract

Cooking robots have long been desired by the commercial market, while the technical challenge is still significant. A major difficulty comes from the demand of perceiving and handling liquid with different properties. This paper presents a robot system that mixes batter and makes pancakes out of it, where understanding and handling the viscous liquid is an essential component. The system integrates Haptic Sensing and control algorithms to autonomously stir flour and water to achieve the desired batter uniformity, estimate the batter's properties such as the water-flour ratio and liquid level, as well as perform precise manipulations to pour the batter into any specified shape. Experimental results show the system's capability to always produce batter of desired uniformity, estimate water-flour ratio and liquid level precisely, and accurately pour it into complex shapes. This research showcases the potential for robots to assist in kitchens and step towards commercial culinary automation.

Figures (10)

• Figure 1: We developed an intelligent pancake-making system to perform batter stirring and pouring. Our system can perceive batter and adjust the pouring policy accordingly.
• Figure 2: Pipeline for our pancake-making system. In the preparation phase, the robot grabs the whisk and then measures the precise position and size of the mixing bowl. Then follows the preliminary stir phase and perceptive stirring until the batter is uniform. Then the robot measures batter's liquid level and water-flour ratio. For pouring, we first plan the trajectory of the robot, then use a stroke control model with liquid level and water-flour ratio as input to regulate the robot's moving speed to maintain the desired stroke width.
• Figure 3: Batter uniformity analysis. (a) Reduction in resistance torque due to increasing uniformity during the stirring of five batches of batter, with water-flour ratios ranging from $1.1$ to $1.5$. Black crosses mark the points of uniformity as assessed by human observers, who use a scale ranging from $-3$ to $3$: (b) non-uniform example($-3$), (c) uniform example($0$), and (d) over-uniform example($3$).
• Figure 4: This figure shows the pipeline of liquid level and water-flour ratio estimation. (a) The trajectory of the pushing motion of each trial is represented by blue lines. Torque measurements are acquired from each pushing motion, resulting in a Torque-Distance to bottom curve (b). This curve is used to estimate liquid level. Following this, a Torque-Immersion curve (c) is derived from the estimated liquid level. Finally, the water-flour ratio is estimated using this data through our data-driven model (d).
• Figure 5: Viusal Feedback Control Pipeline. When the vertical distance between the top and bottom pixels of the batter segment starts to increase, it means the batter begins to flow onto the spout. When the distance stops to increase, it signals the batter has flowed to the end of the spout.